Mbed test linker Errors - IAR8

[deepikabhavnani]

Mbed test build fails on IAR8 feature branch with below error, refer PR #5329 for more information:

        Link: heap_and_stack
        [DEBUG] Link: /usr/local/iccarm-8.20.1.57/bin/ilinkarm -f /builds/ws/mbed-os-build-matrix/341/KL25Z_IAR8/sources/mbed-os/BUILD/tests/KL25Z/IAR/./TESTS/mbedmicro-rtos-mbed/heap_and_stack/.link_files.txt
        [DEBUG] Return: 2
        [DEBUG] Output: 
        [DEBUG] Output:    IAR ELF Linker V8.20.1.57/LNX for ARM
        [DEBUG] Output:    Copyright 2007-2017 IAR Systems AB.
        [DEBUG] Output: 
        [DEBUG] Output:   32 576 bytes of readonly  code memory
        [DEBUG] Output:    1 880 bytes of readonly  data memory
        [DEBUG] Output:   16 243 bytes of readwrite data memory
        [DEBUG] Output: 
        [DEBUG] Output: Errors: 1
        [DEBUG] Output: Warnings: none
        [DEBUG] Output: 
        [DEBUG] Output: Link time:   0.17 (CPU)   1.00 (elapsed)
        [DEBUG] Errors: Error[Lp011]: section placement failed
        [DEBUG] Errors:             unable to place sections/blocks with a total estimated minimum size
        [DEBUG] Errors:                       of 0xf4 bytes (the smallest of which is Section
        [DEBUG] Errors:                       .bss.os.msgqueue.cb (rtx_lib.o #11) (min size 0x34, align
        [DEBUG] Errors:                       0x4) in: <[0x000000c0-0x000003ff],
        [DEBUG] Errors:                       [0x00008920-0x0001ffff], [0x1ffff31c-0x1ffff31f],
        [DEBUG] Errors:                       [0x200029c0-0x20002fff]>

Description

• Type: Bug
• Priority: Blocker

Target
KL25Z and KL46Z

Toolchain:
IAR8.2

[0xc0170]

Is this included also with dynamic heap support for IAR8 ? I assume not.

[deepikabhavnani]

It is with dynamic heap support. Got the issue on dynamic heap PR

[deepikabhavnani]

Splited RAM section not compatible with dynamic HEAP. RAM define/section if splited, IAR expanding size considers just one section while heap creation which results in smaller chunk for heap.
Fixed in commit 8e57107

[deepikabhavnani]

0xc0170 - Please add Freescale team for review

[0xc0170]

Fixed in commit 8e57107

Is it functional ? I recall when we had one region (some time ago), allocation need to be aware about boundaries (no object should cross the boundary that is set to 0x2000 0000 as I recall. Limitation of their controller that does not support access independently to both RAM banks at the same time). I can not find the issue now for reference :/ That fix is the same for ARMCC and GCC, one RAM region?

@mmahadevan108 @maclobdell Can you please have a look at this issue.

[deepikabhavnani]

Is it functional ?

We did the morph test on #5329, after fix but we do not have any KL series device in CI testing.

Limitation of their controller

As per @c1728p9, crossing boundaries should not be an issue with M0 devices. I am not sure of any controller limitation, hence would be best to have review from Freescale.

That fix is the same for ARMCC and GCC, one RAM region

Fix is applied only for IAR (dynamic heap changes). No change in ARMCC and GCC linker scripts

[mmahadevan108]

Looks good to me.

[ciarmcom]

ARM Internal Ref: MBOTRIAGE-140

[deepikabhavnani]

This issue was resolved by IAR in upcoming release.