Lpc1768 bootloader support #5038

chrissnow · 2017-09-06T14:06:18Z

Description

Enable bootloaders for LPC1768 targets.

CRP word was being added to application builds which would overlap the bootloader and cause an sector erase inside the bootloader region.

Linker scripts have been updated.

Status

READY

Migrations

NO

Related PRs

Todos

Deploy notes

Steps to test or reproduce

0xc0170 · 2017-09-06T14:21:57Z

CRP word was being added to application builds which would overlap the bootloader and cause an sector erase inside the bootloader region.

How is this to be done? Is it via targets/TARGET_NXP/TARGET_LPC176X/device/CRP.c the file that places the number directly at defined memory?

chrissnow · 2017-09-06T15:07:39Z

It was previously inside the startup file but since the assembly flags didn't provide anything to distinguish from a bootloaded application to a normal one I moved it to CRP.c

mbed-os/targets/TARGET_NXP/TARGET_LPC176X/device/TOOLCHAIN_ARM_STD/startup_LPC17xx.S

Line 86 in 2f2da77

AREA |.ARM.__at_0x02FC|, CODE, READONLY

I now realise there is some more work to do as I forgot there is a startup file per toolchain.

chrissnow · 2017-09-06T15:11:07Z

@0xc0170 do you think this is the right approach before I sort the others?

0xc0170 · 2017-09-07T10:47:54Z

Jenkins CI fails:

21:59:11 [ARCH_PRO_arm-none-eabi-gcc_minimal] /usr/local/gcc-arm-none-eabi-4_9-2015q3/bin/../lib/gcc/arm-none-eabi/4.9.3/../../../../arm-none-eabi/bin/ld:.build/ARCH_PRO_arm-none-eabi-gcc_minimal/.link_script.ld:15: warning: memory region `m_data' not declared
21:59:11 [ARCH_PRO_arm-none-eabi-gcc_minimal] /usr/local/gcc-arm-none-eabi-4_9-2015q3/bin/../lib/gcc/arm-none-eabi/4.9.3/../../../../arm-none-eabi/bin/ld: section .CRPSection loaded at [000002fc,000002ff] overlaps section .text loaded at [00000000,00013263]
21:59:11 [ARCH_PRO_arm-none-eabi-gcc_minimal] collect2: error: ld returned 1 exit status

@0xc0170 do you think this is the right approach before I sort the others?

I do not fully understand the since the assembly flags didn't provide anything to distinguish from a bootloaded application to a normal one

chrissnow · 2017-09-07T11:14:38Z

I have messed up the GCC linker script, odd how it built for me but I know why it's failed.

The problem is I need to know whether to include the CRP word based on a compiler flag such as BOOTLOADER_ADDR, it must not be included in the build if the app is bootloaded (the bootloader already set's CRP)

However that inst available in assembly compilation so I moved it into a C file.
this means the CRP word must be located correctly but GCC doesn't seem to allow static positioning like this

__attribute__((section(".ARM.__at_0x000002FC")))

so I need to place it with the linker.

chrissnow · 2017-09-07T15:11:03Z

CRP is now in the correct place for GCC,IAR and ARM, this will be included in the bootloader and the application.

This is done so if the application bin was flashed directly it would still have a valid CRP.

CRP verified by checking .bin files manually.

b-kelder · 2017-09-12T14:04:41Z

Thanks for getting this to work, I had some trouble figuring it out myself. I tested the implementation and everything is working on my end.

chrissnow · 2017-09-12T14:46:16Z

@bramkelder Glad it works for you too, IAP is fixed in anther PR if you need that too.
you can also use this branch which has it all merged for testing
https://github.com/chrissnow/mbed-os/tree/CSDEV

It's been a bit of trouble to get it all to work!

I think this is now complete.

0xc0170 · 2017-09-19T08:25:03Z

cc @c1728p9

c1728p9 · 2017-09-20T15:09:33Z

targets/TARGET_NXP/TARGET_LPC176X/device/CRP.c

+#elif defined (__ARMCC_VERSION)
+    const long CRP_Key __attribute__((used)) __attribute__((at(APPLICATION_ADDR+0x000002FC))) = CRP;
+#elif defined (__GNUC__)
+    const long CRP_Key __attribute__((used)) __attribute__((section(".CRPSection"))) = CRP;


You could simplify this by handling placement in the linker scripts. This could then be simplified to something like this for all the toolchains:

MBED_SECTION(".CRPSection") const long CRP_Key = CRP;

The scatter file changes would look something like this:

ER_IROM0 MBED_APP_START 0x2FC { ; load address = execution address *.o (RESET, +First) .ANY (+RO) } ER_CRP (MBED_APP_START + 0x2FC) 4) { *.o (.CRPSection) } ER_IROM1 (MBED_APP_START + (0x2FC + 4)) (MBED_APP_SIZE - (0x2FC + 4)) { *(InRoot$$Sections) .ANY (+RO) }

@c1728p9 I tried your suggestion and get the following

[ERROR] Error: L6788E: Scatter-loading of execution region ER_IROM1 to execution address [0x00010300,0x0001737c) will cause the contents of execution region ER_IROM1 at load-address [0x000100c4,0x00017140) to be corrupted at run-time. Error: L6202E: __main.o(!!!main) cannot be assigned to non-root region 'ER_IROM1' Error: L6202E: __scatter.o(!!!scatter) cannot be assigned to non-root region 'ER_IROM1' Error: L6202E: __dczerorl.o(!!dczerorl) cannot be assigned to non-root region 'ER_IROM1' Error: L6202E: __scatter_copy.o(!!handler_copy) cannot be assigned to non-root region 'ER_IROM1' Error: L6202E: __scatter_zi.o(!!handler_zi) cannot be assigned to non-root region 'ER_IROM1' Error: L6202E: anon$$obj.o(Region$$Table) cannot be assigned to non-root region 'ER_IROM1' Error: L6203E: Entry point (0x00010301) lies within non-root region ER_IROM1.

Any assistance appreciated, I'm no expert on linker files!

From what I can tell 'InRoot$$Sections' can only be placed in a root section, as it needs to place some ARM library stuff there. The root section here would be ER_IROM0, instead of ER_IROM1.
Moving it over might solve the issue?

Then again, I have no expert knowledge on this either.

That gets rid of most errors but leaves me with

ER_IROM0 MBED_APP_START 0x2FC { ; load address = execution address *.o (RESET, +First) *(InRoot$$Sections) .ANY (+RO) } ER_CRP (MBED_APP_START + 0x2FC) 4 { *.o (.CRPSection) } ER_IROM1 (MBED_APP_START + (0x2FC + 4)) (MBED_APP_SIZE - (0x2FC + 4)) { .ANY (+RO) }

Error: L6788E: Scatter-loading of execution region ER_IROM1 to execution address
0x00010300,0x00017d80) will cause the contents of execution region ER_IROM1 at load-address [0x000101a4,0x00017c24) to be corrupted at run-time.

It appears to be overwriting itself, though I can't really see how...

This occurs when scatter-loading takes place and an execution region is put in a position where it partially or wholly overwrites another execution region (which can be itself or another region).
For example, the following code generates this error:

LOAD_ROM 0x0000 0x4000 { EXEC1 0x4000 0x4000 { *(+RW,+ZI) } EXEC2 0x0000 0x4000 { *(+RO) } }

This code does not generate the error:

LOAD_ROM 0x0000 0x4000 { EXEC1 0x0000 0x4000 { * (+RO) } EXEC2 0x4000 0x4000 { * (+RW,+ZI) } }

Source: https://developer.arm.com/docs/dui0807/f/linker-errors-and-warnings/list-of-the-armlink-error-and-warning-messages

Try adding the FIXED keyword to ER_CRP. This will force it to be at the right address

ER_CRP (MBED_APP_START + 0x2FC) FIXED 4 { *.o (.CRPSection) }

It doesn't seem to be related to the CRP section as removing it still causes

Error: L6788E: Scatter-loading of execution region ER_IROM1 to execution address [0x00010300,0x00017d80) will cause the contents of execution region ER_IROM1 at load-address [0x000101a4,0x00017c24) to be corrupted at run-time.

You can try and use AssignExpr() to make sure ER_IROM 1's execution address cannot overlap with its load address, though I don't know if it will make a difference.

ER_IROM0 MBED_APP_START 0x2FC { ; load address = execution address *.o (RESET, +First) *(InRoot$$Sections) .ANY (+RO) } ER_CRP (MBED_APP_START + 0x2FC) 4 { *.o (.CRPSection) } ER_IROM1 AlignExpr(+0, 0x2000)( (MBED_APP_SIZE - (0x2FC + 4)) { ; Align it at the end of the previous section but set its execution address to 0x20000 .ANY (+RO) }

The code above obviously needs a better address then the hardcoded 0x20000 but it might work in your situation.

Try this:

ER_IROM0 MBED_APP_START 0x2FC { ; load address = execution address *.o (RESET, +First) .ANY (+RO) } ER_CRP (MBED_APP_START + 0x2FC) FIXED 4 { *.o (.CRPSection) } ER_IROM1 (MBED_APP_START + (0x2FC + 4)) FIXED (MBED_APP_SIZE - (0x2FC + 4)) { *(InRoot$$Sections) .ANY (+RO) }

0xc0170 · 2017-09-29T13:46:21Z

@chrissnow Have you find a way to simplify the crc placement, or staying as it is now here?

chrissnow · 2017-09-29T14:44:48Z

@0xc0170 I'd like to simplify it but still need to sort out the scatter file issue as above.
Any suggestions welcome :-)

chrissnow · 2017-09-29T22:06:18Z

@0xc0170 I think it's going to have to go as is unless @c1728p9 or @bramkelder can get the arm linker to work.

I appreciate it's not as elegant as it could be but I seem to be fighting a losing battle here!

…piled

CRP value can be set through a macro in mbed_app such as "macros": [ "CRP=CRP_NONE" ]

chrissnow · 2017-09-30T21:39:41Z

@0xc0170 I'm pretty sure this is now finished, Thanks @c1728p9 for figuring the linker issues out :-)

I have no idea what CircleCI is but it's error seems unrelated and common to other recent PR CI run's

0xc0170 · 2017-10-02T09:59:03Z

/morph test

mbed-bot · 2017-10-02T12:41:28Z

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph test

Output

mbed Build Number: 1490

All builds and test passed!

theotherjimmy · 2017-10-02T15:20:54Z

@chrissnow Could you change the sha of the last commit so that it kicks out circle CI? We'll rerun testing after that.

chrissnow · 2017-10-02T18:35:31Z

@theotherjimmy That should have done it.

theotherjimmy · 2017-10-02T18:37:49Z

/morph test

theotherjimmy · 2017-10-02T18:37:53Z

Thanks!

mbed-bot · 2017-10-03T07:07:23Z

Result: ABORTED

Your command has finished executing! Here's what you wrote!

/morph test

0xc0170 · 2017-10-04T08:18:07Z

/morph test

mbed-bot · 2017-10-04T18:54:11Z

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph test

Output

mbed Build Number: 1521

All builds and test passed!

0xc0170 added the needs: review label Sep 6, 2017

0xc0170 approved these changes Sep 19, 2017

View reviewed changes

0xc0170 added needs: CI and removed needs: review labels Sep 19, 2017

chrissnow mentioned this pull request Sep 20, 2017

ARMCC - FromELF multiple regions #5021

Closed

c1728p9 reviewed Sep 20, 2017

View reviewed changes

0xc0170 added needs: work and removed needs: CI labels Sep 29, 2017

chrissnow added 5 commits September 30, 2017 19:00

Enable LPC1768 bootloader support

fc503b8

Move CRP out of startup and into CRP.c so it can be conditionally com…

a08fc2b

…piled

Linker update for bootloader support

f8f5483

Ensure CRP is set correctly for IAR, GCC and ARM

d640472

CRP value can be set through a macro in mbed_app such as "macros": [ "CRP=CRP_NONE" ]

Implement MBED_USED

fb27e71

chrissnow force-pushed the LPC1768-Bootloader branch from 636c1b3 to 0070c78 Compare September 30, 2017 18:28

b-kelder approved these changes Oct 2, 2017

View reviewed changes

0xc0170 added needs: CI and removed needs: work labels Oct 2, 2017

Simplify CRP placement.

82ae53a

chrissnow force-pushed the LPC1768-Bootloader branch from 0070c78 to 82ae53a Compare October 2, 2017 18:30

0xc0170 added ready for merge and removed needs: CI labels Oct 4, 2017

theotherjimmy merged commit a0b624b into ARMmbed:master Oct 5, 2017

theotherjimmy added release-version: 5.6.3 and removed ready for merge labels Oct 5, 2017

chrissnow mentioned this pull request Oct 13, 2017

LPC1769 port #5025

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Lpc1768 bootloader support #5038

Lpc1768 bootloader support #5038

chrissnow commented Sep 6, 2017

0xc0170 commented Sep 6, 2017

chrissnow commented Sep 6, 2017

chrissnow commented Sep 6, 2017

0xc0170 commented Sep 7, 2017

chrissnow commented Sep 7, 2017

chrissnow commented Sep 7, 2017

b-kelder commented Sep 12, 2017 •

edited

Loading

chrissnow commented Sep 12, 2017

0xc0170 commented Sep 19, 2017

c1728p9 Sep 20, 2017

chrissnow Sep 25, 2017

b-kelder Sep 26, 2017

chrissnow Sep 26, 2017

b-kelder Sep 26, 2017

c1728p9 Sep 26, 2017

chrissnow Sep 26, 2017

b-kelder Sep 27, 2017

c1728p9 Sep 29, 2017

0xc0170 commented Sep 29, 2017

chrissnow commented Sep 29, 2017

chrissnow commented Sep 29, 2017

chrissnow commented Sep 30, 2017

0xc0170 commented Oct 2, 2017

mbed-bot commented Oct 2, 2017

theotherjimmy commented Oct 2, 2017

chrissnow commented Oct 2, 2017

theotherjimmy commented Oct 2, 2017

theotherjimmy commented Oct 2, 2017

mbed-bot commented Oct 3, 2017

0xc0170 commented Oct 4, 2017

mbed-bot commented Oct 4, 2017

Lpc1768 bootloader support #5038

Lpc1768 bootloader support #5038

Conversation

chrissnow commented Sep 6, 2017

Description

Status

Migrations

Related PRs

Todos

Deploy notes

Steps to test or reproduce

0xc0170 commented Sep 6, 2017

chrissnow commented Sep 6, 2017

chrissnow commented Sep 6, 2017

0xc0170 commented Sep 7, 2017

chrissnow commented Sep 7, 2017

chrissnow commented Sep 7, 2017

b-kelder commented Sep 12, 2017 • edited Loading

chrissnow commented Sep 12, 2017

0xc0170 commented Sep 19, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

0xc0170 commented Sep 29, 2017

chrissnow commented Sep 29, 2017

chrissnow commented Sep 29, 2017

chrissnow commented Sep 30, 2017

0xc0170 commented Oct 2, 2017

mbed-bot commented Oct 2, 2017

Output

theotherjimmy commented Oct 2, 2017

chrissnow commented Oct 2, 2017

theotherjimmy commented Oct 2, 2017

theotherjimmy commented Oct 2, 2017

mbed-bot commented Oct 3, 2017

0xc0170 commented Oct 4, 2017

mbed-bot commented Oct 4, 2017

Output

b-kelder commented Sep 12, 2017 •

edited

Loading