[PATCH 1/1] sandbox: allow building sandbox_spl with CONFIG_DEBUG

[PATCH 1/1] sandbox: allow building sandbox_spl with CONFIG_DEBUG

[Heinrich Schuchardt]

Building sandbox_spl with CONFIG_DEBUG leads to errors due to missing
symbols:

    /usr/bin/ld: common/spl/spl_fit.o: in function `spl_fit_upload_fpga':
    common/spl/spl_fit.c:595: undefined reference to `fpga_load'
    /usr/bin/ld: test/test-main.o: in function `dm_test_post_run':
    test/test-main.c:124: undefined reference to `crc8'
    /usr/bin/ld: test/test-main.o: in function `dm_test_pre_run':
    test/test-main.c:95: undefined reference to `crc8'
    collect2: error: ld returned 1 exit status

This is due to -Og not eliminating unused functions.

Add FPGA and CRC8 support to the defconfig. Sandbox tests for
SPL_FPGA and CRC8 should be created. So enabling these setting
is advised anyway.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
---
 configs/sandbox_spl_defconfig | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/configs/sandbox_spl_defconfig b/configs/sandbox_spl_defconfig
index 4e0021a76f..851c3b687a 100644
--- a/configs/sandbox_spl_defconfig
+++ b/configs/sandbox_spl_defconfig
@@ -33,6 +33,7 @@ CONFIG_SPL_NO_BSS_LIMIT=y
 CONFIG_HANDOFF=y
 CONFIG_SPL_BOARD_INIT=y
 CONFIG_SPL_ENV_SUPPORT=y
+CONFIG_SPL_FPGA=y
 CONFIG_SPL_I2C=y
 CONFIG_SPL_RTC=y
 CONFIG_CMD_CPU=y
@@ -126,6 +127,8 @@ CONFIG_DM_DEMO=y
 CONFIG_DM_DEMO_SIMPLE=y
 CONFIG_DM_DEMO_SHAPE=y
 CONFIG_SPL_FIRMWARE=y
+CONFIG_DM_FPGA=y
+CONFIG_SANDBOX_FPGA=y
 CONFIG_GPIO_HOG=y
 CONFIG_QCOM_PMIC_GPIO=y
 CONFIG_SANDBOX_GPIO=y
@@ -237,6 +240,7 @@ CONFIG_FS_CRAMFS=y
 CONFIG_CMD_DHRYSTONE=y
 CONFIG_RSA_VERIFY_WITH_PKEY=y
 CONFIG_TPM=y
+CONFIG_SPL_CRC8=y
 CONFIG_LZ4=y
 CONFIG_ZSTD=y
 CONFIG_ERRNO_STR=y
-- 
2.38.1

Re: [PATCH 1/1] sandbox: allow building sandbox_spl with CONFIG_DEBUG

[Simon Glass]

On Sat, 18 Feb 2023 at 01:34, Heinrich Schuchardt
<heinrich.schuchardt@canonical.com> wrote:
>
> Building sandbox_spl with CONFIG_DEBUG leads to errors due to missing
> symbols:
>
>     /usr/bin/ld: common/spl/spl_fit.o: in function `spl_fit_upload_fpga':
>     common/spl/spl_fit.c:595: undefined reference to `fpga_load'
>     /usr/bin/ld: test/test-main.o: in function `dm_test_post_run':
>     test/test-main.c:124: undefined reference to `crc8'
>     /usr/bin/ld: test/test-main.o: in function `dm_test_pre_run':
>     test/test-main.c:95: undefined reference to `crc8'
>     collect2: error: ld returned 1 exit status
>
> This is due to -Og not eliminating unused functions.
>
> Add FPGA and CRC8 support to the defconfig. Sandbox tests for
> SPL_FPGA and CRC8 should be created. So enabling these setting
> is advised anyway.
>
> Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
> ---
>  configs/sandbox_spl_defconfig | 4 ++++
>  1 file changed, 4 insertions(+)
>

Reviewed-by: Simon Glass <sjg@chromium.org>

Re: [PATCH 1/1] sandbox: allow building sandbox_spl with CONFIG_DEBUG

[Eugen Hristev]

On 2/21/23 21:35, Simon Glass wrote:
> On Sat, 18 Feb 2023 at 01:34, Heinrich Schuchardt
> <heinrich.schuchardt@canonical.com> wrote:
>>
>> Building sandbox_spl with CONFIG_DEBUG leads to errors due to missing
>> symbols:
>>
>>      /usr/bin/ld: common/spl/spl_fit.o: in function `spl_fit_upload_fpga':
>>      common/spl/spl_fit.c:595: undefined reference to `fpga_load'
>>      /usr/bin/ld: test/test-main.o: in function `dm_test_post_run':
>>      test/test-main.c:124: undefined reference to `crc8'
>>      /usr/bin/ld: test/test-main.o: in function `dm_test_pre_run':
>>      test/test-main.c:95: undefined reference to `crc8'
>>      collect2: error: ld returned 1 exit status
>>
>> This is due to -Og not eliminating unused functions.
>>
>> Add FPGA and CRC8 support to the defconfig. Sandbox tests for
>> SPL_FPGA and CRC8 should be created. So enabling these setting
>> is advised anyway.
>>
>> Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
>> ---
>>   configs/sandbox_spl_defconfig | 4 ++++
>>   1 file changed, 4 insertions(+)
>>
> 
> Reviewed-by: Simon Glass <sjg@chromium.org>

Hello Heinrich, Simon,

I am facing similar issues on different defconfig when enabling 
OPTIMIZE_DEBUG.
I don't think enabling SPL_FPGA and DM_FPGA is the solution there.
Do you know why is the drivers/fpga/fpga.c built all the time, with no 
Kconfig selected ?
I think that maybe spl_fit_upload_fpga and fpga_load should be built 
conditionally , I would like to have SPL load a FIT, but not anything 
related to FPGA in my config.

Thanks,
Eugen

Re: [PATCH 1/1] sandbox: allow building sandbox_spl with CONFIG_DEBUG

[Simon Glass]

Hi Eugen,

On Fri, 24 Feb 2023 at 06:59, Eugen Hristev <eugen.hristev@collabora.com> wrote:
>
> On 2/21/23 21:35, Simon Glass wrote:
> > On Sat, 18 Feb 2023 at 01:34, Heinrich Schuchardt
> > <heinrich.schuchardt@canonical.com> wrote:
> >>
> >> Building sandbox_spl with CONFIG_DEBUG leads to errors due to missing
> >> symbols:
> >>
> >>      /usr/bin/ld: common/spl/spl_fit.o: in function `spl_fit_upload_fpga':
> >>      common/spl/spl_fit.c:595: undefined reference to `fpga_load'
> >>      /usr/bin/ld: test/test-main.o: in function `dm_test_post_run':
> >>      test/test-main.c:124: undefined reference to `crc8'
> >>      /usr/bin/ld: test/test-main.o: in function `dm_test_pre_run':
> >>      test/test-main.c:95: undefined reference to `crc8'
> >>      collect2: error: ld returned 1 exit status
> >>
> >> This is due to -Og not eliminating unused functions.
> >>
> >> Add FPGA and CRC8 support to the defconfig. Sandbox tests for
> >> SPL_FPGA and CRC8 should be created. So enabling these setting
> >> is advised anyway.
> >>
> >> Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
> >> ---
> >>   configs/sandbox_spl_defconfig | 4 ++++
> >>   1 file changed, 4 insertions(+)
> >>
> >
> > Reviewed-by: Simon Glass <sjg@chromium.org>
>
> Hello Heinrich, Simon,
>
> I am facing similar issues on different defconfig when enabling
> OPTIMIZE_DEBUG.
> I don't think enabling SPL_FPGA and DM_FPGA is the solution there.
> Do you know why is the drivers/fpga/fpga.c built all the time, with no
> Kconfig selected ?

That's because the fpga/ directory itself is guarded.

> I think that maybe spl_fit_upload_fpga and fpga_load should be built
> conditionally , I would like to have SPL load a FIT, but not anything
> related to FPGA in my config.

Sounds right. Overall the FPGA subsystem could use some work, i.e. a
proper uclass API with methods, etc. There is discussion about it
elsewhere on the mailing list.

Regards,
Simon

Re: [PATCH 1/1] sandbox: allow building sandbox_spl with CONFIG_DEBUG

[Simon Glass]

Hi Eugen,

On Fri, 24 Feb 2023 at 06:59, Eugen Hristev <eugen.hristev@collabora.com> wrote:
>
> On 2/21/23 21:35, Simon Glass wrote:
> > On Sat, 18 Feb 2023 at 01:34, Heinrich Schuchardt
> > <heinrich.schuchardt@canonical.com> wrote:
> >>
> >> Building sandbox_spl with CONFIG_DEBUG leads to errors due to missing
> >> symbols:
> >>
> >>      /usr/bin/ld: common/spl/spl_fit.o: in function `spl_fit_upload_fpga':
> >>      common/spl/spl_fit.c:595: undefined reference to `fpga_load'
> >>      /usr/bin/ld: test/test-main.o: in function `dm_test_post_run':
> >>      test/test-main.c:124: undefined reference to `crc8'
> >>      /usr/bin/ld: test/test-main.o: in function `dm_test_pre_run':
> >>      test/test-main.c:95: undefined reference to `crc8'
> >>      collect2: error: ld returned 1 exit status
> >>
> >> This is due to -Og not eliminating unused functions.
> >>
> >> Add FPGA and CRC8 support to the defconfig. Sandbox tests for
> >> SPL_FPGA and CRC8 should be created. So enabling these setting
> >> is advised anyway.
> >>
> >> Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
> >> ---
> >>   configs/sandbox_spl_defconfig | 4 ++++
> >>   1 file changed, 4 insertions(+)
> >>
> >
> > Reviewed-by: Simon Glass <sjg@chromium.org>
>
> Hello Heinrich, Simon,
>
> I am facing similar issues on different defconfig when enabling
> OPTIMIZE_DEBUG.
> I don't think enabling SPL_FPGA and DM_FPGA is the solution there.
> Do you know why is the drivers/fpga/fpga.c built all the time, with no
> Kconfig selected ?

That's because the fpga/ directory itself is guarded.

> I think that maybe spl_fit_upload_fpga and fpga_load should be built
> conditionally , I would like to have SPL load a FIT, but not anything
> related to FPGA in my config.

Sounds right. Overall the FPGA subsystem could use some work, i.e. a
proper uclass API with methods, etc. There is discussion about it
elsewhere on the mailing list.

Regards,
Simon

Applied to u-boot-dm, thanks!