Hi Bohdan,
Sorry I’m not an expert of linker scripts. Here’s my understanding.
For GCC, as long as you add the item in the .copy.table, the toolchain would generate corresponding code to copy codes and data.
For Armclang, you don’t even need the copy table. The toolchain does the copy as long as the addresses of the “load region (LR)” and “execution region (ER)” are different.
For IAR, I have no knowledge of it.
One thing I noticed is that GCC picks up the (.ram_vectors) into the ram vector section while Armclang pick the
* (RESET_RAM).
i
Best Regards,
Kevin
From: Bohdan.Hunko@infineon.com <Bohdan.Hunko@infineon.com>
Sent: Wednesday, July 6, 2022 3:01 AM
To: Kevin Peng <Kevin.Peng@arm.com>; tf-m@lists.trustedfirmware.org
Subject: RE: Automatically copy RAM_VECTORS when using ARMClang and IAR
Hi Kevin,
So the situation is as follows:
Therefore it is a bit inconvenient to copy .ram_vectors manually for IAR and Clang
As for now ER_RAM_VECTORS is copied in startup file for Clang and IAR manually and for GCC automatically. So currently there is no bug and everything works as expected.
But I am trying to understand whether there is a way to change Clang and IAR startups so
ER_RAM_VECTORS will be copied automatically for these compilers too.
So far I have not been able to find such a way, so I will appreciate if someone knows how to do this. But if no one knows the solution we can leave it as is.
Regards,
Bohdan Hunko
Cypress Semiconductor Ukraine
Engineer
CSUKR CSS ICW SW FW
Mobile: +38099 50 19 714
Bohdan.Hunko@infineon.com
From: Kevin Peng <Kevin.Peng@arm.com>
Sent: Tuesday, 28 June 2022 06:20
To: Hunko Bohdan (CSUKR CSS ICW SW FW) <Bohdan.Hunko@infineon.com>;
tf-m@lists.trustedfirmware.org
Subject: RE: Automatically copy RAM_VECTORS when using ARMClang and IAR
|
Caution: This e-mail originated outside Infineon Technologies. Do not click on links or open
attachments unless you
validate it is safe. |
Hi Bohdan,
I don’t know about IAR.
For Armclang, I think __main does copy codes and data:
“__main performs code and data copying, decompression, and zero initialization of the ZI data.”
I had a quick look, so probably I’m wrong.
Best Regards,
Kevin
From: Bohdan.Hunko--- via TF-M <tf-m@lists.trustedfirmware.org>
Sent: Monday, June 27, 2022 10:02 PM
To: tf-m@lists.trustedfirmware.org
Subject: [TF-M] Re: Automatically copy RAM_VECTORS when using ARMClang and IAR
Hi,
I would highly appreciate any help on this one.
Thanks!
Regards,
Bohdan Hunko
Cypress Semiconductor Ukraine
Engineer
CSUKR CSS ICW SW FW
Mobile: +38099 50 19 714
Bohdan.Hunko@infineon.com
From: Bohdan.Hunko--- via TF-M <tf-m@lists.trustedfirmware.org>
Sent: Wednesday, 13 April 2022 21:46
To: tf-m@lists.trustedfirmware.org
Subject: [TF-M] Automatically copy RAM_VECTORS when using ARMClang and IAR
|
Caution: This e-mail originated outside Infineon Technologies. Do not click on links or open
attachments unless you
validate it is safe. |
Hi,
When poking around some startup files I have found interesting place related to RAM_VECTORS support
CMSIS have __PROGRAM_START macro which is different for each compiler.
For GCC it uses __cmsis_start, for ARMClang - __main and for IAR - __iar_program_start
Basically each of the functions should copy several sections (.TFM_DATA for example) from FLASH to RAM and zero out some parts of RAM (for .TFM_BSS for example)
In current implementation GCC __cmsis_start function also copies the vector table from FLASH to SRAM (if RAM_VECTORS are enabled)
But ARMClang and IAR equivalents of that function (__main, __iar_program_start) does not seem to take care of copying vector table, so platforms startup should do that
I wonder if there is a way to change linker script in a way which will make copying of vector table automatic (by compiler dependent function).
This will make platform startups a bit cleaner and will allow platform to just use __PROGRAM_START macro without any additional code to copy vector table.
From what I see IAR has “initialize by copy” syntaxis so I think it may be used to tell IAR to automatically copy vector table.
It is a bit more tricky with ARMClang as I have not found a way to do that there.
I am not a big expert in ARMClang and IAR so maybe someone may help me here, give some directions or confirm that currently there is no way to make this idea work.
Basically the intention is to simplify platform startup code and offload common operations to compiler specific platform independent functions.
Regards,
Bohdan Hunko
Cypress Semiconductor Ukraine
Engineer
CSUKR CSS ICW SW FW
Mobile: +38099 50 19 714
Bohdan.Hunko@infineon.com