nrf52810_pca10040 SRAM space not enough

[wentongwu]

Describe the bug
it seems nrf52810_pca10040 SRAM space is not enough for current version. It fails on link stage.

To Reproduce
Steps to reproduce the behavior:

1. mkdir build; cd build
2. cmake -GNinja -DBOARD=nrf52810_pca10040 ..
3. ninja
4. See error

Screenshots or console output

for case tests/kernel/sched/preempt on platform nrf52810_pca10040
Memory region Used Size Region Size %age Used
FLASH: 63104 B 192 KB 32.10%
SRAM: 25 KB 24 KB 104.17%
IDT_LIST: 56 B 2 KB 2.73/opt/zephyr-sdk-0.10/arm-zephyr-eabi/bin/../lib/gcc/arm-zephyr-eabi/8.3.0/../../../../arm-zephyr-eabi/bin/ld: zephyr/zephyr_prebuilt.elf section priv_stacks_noinit' will not fit in region SRAM'
/opt/zephyr-sdk-0.10/arm-zephyr-eabi/bin/../lib/gcc/arm-zephyr-eabi/8.3.0/../../../../arm-zephyr-eabi/bin/ld: section .intList VMA [0000000020006000,0000000020006037] overlaps section priv_stacks_noinit VMA [0000000020003800,00000000200063ff]
/opt/zephyr-sdk-0.10/arm-zephyr-eabi/bin/../lib/gcc/arm-zephyr-eabi/8.3.0/../../../../arm-zephyr-eabi/bin/ld: region `SRAM' overflowed by 1024 bytes
collect2: error: ld returned 1 exit status

for case tests/kernel/poll on platform nrf52810_pca10040
-sdk-0.10/arm-zephyr-eabi/bin/arm-zephyr-eabi-objdump -S zephyr.elf > zephyr.lst && /opt/zephyr-sdk-0.10/arm-zephyr-eabi/bin/arm-zephyr-eabi-readelf -e zephyr.elf > zephyr.stat
/opt/zephyr-sdk-0.10/arm-zephyr-eabi/bin/../lib/gcc/arm-zephyr-eabi/8.3.0/../../../../arm-zephyr-eabi/bin/ld: zephyr/zephyr.elf section priv_stacks_noinit' will not fit in region SRAM'
/opt/zephyr-sdk-0.10/arm-zephyr-eabi/bin/../lib/gcc/arm-zephyr-eabi/8.3.0/../../../../arm-zephyr-eabi/bin/ld: region `SRAM' overflowed by 2048 bytes

Environment (please complete the following information):

• OS: Ubuntu
• Toolchain (Zephyr SDK, 0.10.0)
• Commit SHA 193fb97

[cinlyooi-intel]

Seeing the same error at 193fb97

[wentongwu]

@andrewboie this is the same issue when PR#15005 run CI test.

[ioannisg]

nrf52810 is a low-end SOC - several tests do not run for this device, anyways.

I am wondering whether it is a problem of increasing the PRIVILEGED_STACK_SIZE to default of 1024.

[wentongwu]

nrf52810 is a low-end SOC - several tests do not run for this device, anyways.
I'm not running cases on that platform, but CI test want to do that when I submit PR #15005.
I am wondering whether it is a problem of increasing the PRIVILEGED_STACK_SIZE to default of 1024.
you mean megered patches increasing PRIVILEGED_STACK_SIZE cause this platform's failure?

[andrewboie]

I am wondering whether it is a problem of increasing the PRIVILEGED_STACK_SIZE to default of 1024.

My guess as well. How many threads does this test create?

[wentongwu]

I am wondering whether it is a problem of increasing the PRIVILEGED_STACK_SIZE to default of 1024.

My guess as well. How many threads does this test create?

from priv_stacks_hash.c, sched/preempt has 11 threads and /kernel/poll has 10 threads.

[wentongwu]

for sched/preempt/, if PRIVILEGED_STACK_SIZE = 256, zephyr_prebuilt.elf is 15872 Bytes, so if set PRIVILEGED_STACK_SIZE = 1024, zephyr_prebuilt.elf will almost occupy the whole sram. Finally it should add kobjects, so when do the linking for zephyr.elf, there isn't enough space.

I see the commit changing priv stack size from 256 to 1024, so the work for proving maximum syscall stack depth is ongoing? @andrewboie

[andrewboie]

@wentongwu I can't reproduce the failure in tests/kernel/sched/preempt, can you retry?

[ 90%] Linking C executable zephyr_prebuilt.elf
Memory region         Used Size  Region Size  %age Used
           FLASH:       62584 B       192 KB     31.83%
            SRAM:         23 KB        24 KB     95.83%
        IDT_LIST:          56 B         2 KB      2.73%
[ 90%] Built target zephyr_prebuilt