M-Labs
/
thermostat
9
0
Fork 5
Code Issues 9 Pull Requests 5 Releases Wiki Activity

software_dfu #46

Merged
sb10q merged 8 commits from software_dfu into master 2021-01-13 11:59:06 +08:00
Conversation 5 Commits 8 Files Changed 5 +29 -30
4 changed files with 29 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files
Showing only changes of commit 2e6116016a - Show all commits

2
README.md
View File

@ -124,7 +124,7 @@ formatted as line-delimited JSON.
| `load [0/1]` | Restore configuration for channel all/0/1 from flash |
| `save [0/1]` | Save configuration for channel all/0/1 to flash |
| `reset` | Reset the device |
| `dfu` | Reset device and enters USB device firmware update (DFU) mode |
| `dfu` | Reset device and enters USB device firmware update (DFU) mode |
sb10q commented 2021-01-12 16:34:25 +08:00
Outdated
Review
Copy Link

Table formatting does not seem OK.

Table formatting does not seem OK.
| `ipv4 <X.X.X.X/L> [Y.Y.Y.Y]` | Configure IPv4 address, netmask length, and optional default gateway |

5
memory.x
View File

@ -3,12 +3,13 @@ MEMORY
FLASH (rx) : ORIGIN = 0x8000000, LENGTH = 1024K
/* reserved for config data */
CONFIG (rx) : ORIGIN = 0x8100000, LENGTH = 16K
RAM (xrw) : ORIGIN = 0x20000000, LENGTH = 111K
RAM (xrw) : ORIGIN = 0x20000000, LENGTH = 112K - 4
/* reserved for DFU trigger message */
DFU_MSG (wrx) : ORIGIN = 0x2001BC00, LENGTH = 1K
DFU_MSG (wrx) : ORIGIN = 0x2001BFFC, LENGTH = 4
sb10q commented 2021-01-12 17:37:57 +08:00
Outdated
Review
Copy Link

The length should be 4 and not 1K AFAICT?

The length should be 4 and not 1K AFAICT?
RAM2 (xrw) : ORIGIN = 0x2001C000, LENGTH = 16K
RAM3 (xrw) : ORIGIN = 0x20020000, LENGTH = 64K
CCMRAM (rw) : ORIGIN = 0x10000000, LENGTH = 64K
}
_stack_start = ORIGIN(CCMRAM) + LENGTH(CCMRAM);
_dfu_msg = ORIGIN(DFU_MSG);

49
src/dfu.rs
View File

@ -1,28 +1,25 @@
use cortex_m_rt::{pre_init};
/// RAM location used to store DFU trigger message
const DFU_MSG_ADDR: usize = 0x2001BC00;
const DFU_TRIG_MSG: u32 = 0xDECAFBAD;
sb10q commented 2021-01-12 16:39:28 +08:00
Outdated
Review
Copy Link

This can and should be automatically obtained via a linker symbol. See this example:
https://git.m-labs.hk/M-Labs/artiq-zynq/src/branch/master/src/runtime/src/kernel/core1.rs#L31-L36
https://git.m-labs.hk/M-Labs/artiq-zynq/src/branch/master/src/runtime/link.x#L10

This can and should be automatically obtained via a linker symbol. See this example: https://git.m-labs.hk/M-Labs/artiq-zynq/src/branch/master/src/runtime/src/kernel/core1.rs#L31-L36 https://git.m-labs.hk/M-Labs/artiq-zynq/src/branch/master/src/runtime/link.x#L10
topquark12 commented 2021-01-12 16:59:47 +08:00
Outdated
Review
Copy Link

The linker currently does not actually do anything to the ram carved out for the dfu trigger message, and I beleive there's no symbol generated for that chunk of memory.

I think I'll have to find the linker script the project is using, fork a copy, generate a symbol for that addr and use it in our project somehow.

The linker currently does not actually do anything to the ram carved out for the dfu trigger message, and I beleive there's no symbol generated for that chunk of memory. I think I'll have to find the linker script the project is using, fork a copy, generate a symbol for that addr and use it in our project somehow.
sb10q commented 2021-01-12 17:02:53 +08:00
Outdated
Review
Copy Link

Indeed, but the linker script can be modified so that such a symbol is generated. This way there isn't any value that needs to be manually synchronized in two places. You may not even need to hardcode the address; the linker could place that variable automatically.

Indeed, but the linker script can be modified so that such a symbol is generated. This way there isn't any value that needs to be manually synchronized in two places. You may not even need to hardcode the address; the linker could place that variable automatically.
sb10q commented 2021-01-12 17:04:05 +08:00
Outdated
Review
Copy Link

Though hardcoding is actually more reliable since after a firmware update the address could have changed otherwise... add a comment that explains that.

Though hardcoding is actually more reliable since after a firmware update the address could have changed otherwise... add a comment that explains that.
sb10q commented 2021-01-12 17:04:35 +08:00
Outdated
Review
Copy Link

The linker script is the second link I gave you.

The linker script is the second link I gave you.
pca006132 commented 2021-01-12 17:09:07 +08:00
Outdated
Review
Copy Link

I think specifying a region that only contains the DFU msg would be reasonable. This way the address would not be changed after update.

I think specifying a region that only contains the DFU msg would be reasonable. This way the address would not be changed after update.
topquark12 commented 2021-01-12 17:10:19 +08:00
Outdated
Review
Copy Link

I thought that linker script is for a zynq? I am not sure where the linker script currently being used is placed, think it is packaged with one of the rust modules.

I thought that linker script is for a zynq? I am not sure where the linker script currently being used is placed, think it is packaged with one of the rust modules.
sb10q commented 2021-01-12 17:11:57 +08:00
Outdated
Review
Copy Link

Okay, you mean the thermostat linker script.
That memory.x you modified can be used, as you can see it already defines the _stack_start symbol.

Okay, you mean the thermostat linker script. That ``memory.x`` you modified can be used, as you can see it already defines the ``_stack_start`` symbol.
topquark12 commented 2021-01-12 17:19:12 +08:00
Outdated
Review
Copy Link

The stack is on CCM, the message is placed in RAM. I can try placing the msg in CCM, not sure what the reset behavior of CCM is yet.

The stack is on CCM, the message is placed in RAM. I can try placing the msg in CCM, not sure what the reset behavior of CCM is yet.
topquark12 commented 2021-01-12 17:22:56 +08:00
Outdated
Review
Copy Link

@pca006132 There's already a dedicated region specified in memory.x . Though the linker does not do anything with that region, and I don't think the linker generates a symbol for that space.

I also haven't found the linker script of the project just yet.

@pca006132 There's already a dedicated region specified in memory.x . Though the linker does not do anything with that region, and I don't think the linker generates a symbol for that space. I also haven't found the linker script of the project just yet.
pca006132 commented 2021-01-12 17:35:08 +08:00
Outdated
Review
Copy Link

The linker script is defined in cortex-m-rt, https://github.com/rust-embedded/cortex-m-rt/blob/master/link.x.in
But I'm not familiar with that, not sure how to add new symbols.

The linker script is defined in `cortex-m-rt`, https://github.com/rust-embedded/cortex-m-rt/blob/master/link.x.in But I'm not familiar with that, not sure how to add new symbols.
sb10q commented 2021-01-12 17:37:22 +08:00
Outdated
Review
Copy Link

@topquark12 what's wrong with simply adding

_dfu_msg = ORIGIN(DFU_MSG)

at the end of thermostat's memory.x?

@topquark12 what's wrong with simply adding ``` _dfu_msg = ORIGIN(DFU_MSG) ``` at the end of thermostat's ``memory.x``?
/// DFU trigger message
const DFU_TRIG_MSG: usize = 0xDECAFBAD;
/// Set DFU trigger
pub unsafe fn trig_dfu() {
let dfu_msg_addr = DFU_MSG_ADDR as *mut usize;
*dfu_msg_addr = DFU_TRIG_MSG;
pub unsafe fn set_dfu_trigger() {
extern "C" {
sb10q commented 2021-01-13 11:07:37 +08:00
Outdated
Review
Copy Link

Why not put it outside the functions so this declaration can be shared?

Why not put it outside the functions so this declaration can be shared?
static mut _dfu_msg: u32;
}
_dfu_msg = DFU_TRIG_MSG;
}
/// Called by reset handler in lib.rs immediately after reset, checks if booting into dfu is needed
/// Called by reset handler in lib.rs immediately after reset.
/// This function should not be called outside of reset handler as
/// bootloader expects MCU to be in reset state when called.
#[pre_init]
unsafe fn __pre_init() {
extern "C" {
sb10q commented 2021-01-13 11:08:24 +08:00
Outdated
Review
Copy Link

see comment above

see comment above
static mut _dfu_msg: u32;
sb10q commented 2021-01-12 16:36:16 +08:00
Outdated
Review
Copy Link

remove blank line

remove blank line
}
let dfu_msg_addr = DFU_MSG_ADDR as *mut usize;
// Check DFU trigger message
if *dfu_msg_addr == DFU_TRIG_MSG{
// Reset message
*dfu_msg_addr = 0x00000000;
if _dfu_msg == DFU_TRIG_MSG {
_dfu_msg = 0x00000000;
sb10q commented 2021-01-12 16:36:38 +08:00
Outdated
Review
Copy Link

space before {

space before {
sb10q commented 2021-01-12 16:36:44 +08:00
Review
Copy Link

remove blank line

remove blank line
// Enable system config controller clock
const RCC_APB2ENR: *mut u32 = 0xE000_ED88 as *mut u32;
@ -41,14 +38,14 @@ unsafe fn __pre_init() {
SYSCFG_MEMRMP,
*SYSCFG_MEMRMP | SYSCFG_MEMRMP_MAP_ROM,
);
// Set stack pointer to bootloader location
asm!("LDR R0, =0x1FFF0000");
asm!("LDR SP,[R0, #0]");
// Jump to bootloader
asm!("LDR R0,[R0, #4]");
asm!("BX R0");
asm!(
// Set stack pointer to bootloader location
"LDR R0, =0x1FFF0000",
"LDR SP,[R0, #0]",
// Jump to bootloader
"LDR R0,[R0, #4]",
"BX R0",
);
}
}
sb10q commented 2021-01-12 16:35:52 +08:00
Outdated
Review
Copy Link

You can put all these lines into a single asm! macro, see https://doc.rust-lang.org/nightly/unstable-book/library-features/asm.html

You can put all these lines into a single ``asm!`` macro, see https://doc.rust-lang.org/nightly/unstable-book/library-features/asm.html

3
src/main.rs
View File

@ -429,8 +429,9 @@ fn main() -> ! {
channels.power_down(i);
}
unsafe {
dfu::trig_dfu();
dfu::set_dfu_trigger();
}
socket.close();
sb10q commented 2021-01-13 11:09:34 +08:00
Outdated
Review
Copy Link

Does this work or do we need to wait for the network driver/hardware to actually send the packet?
What if there are other open sockets? Maybe address this entire issue in a second PR.

Does this work or do we need to wait for the network driver/hardware to actually send the packet? What if there are other open sockets? Maybe address this entire issue in a second PR.
topquark12 commented 2021-01-13 11:13:48 +08:00
Outdated
Review
Copy Link

was just testing out the behavior of that, forgot to remove before committing.

was just testing out the behavior of that, forgot to remove before committing.
SCB::sys_reset();
}
}