Dear David,
In message CBDD3537-16F3-41C8-9929-6C30F56C36DF@phytium.com.cn you wrote:
I am will glad to share my understanding.
Thanks.
I wonder how "adrp" would work in any code that is supposed to be position-independent?
In my understanding, with adrp the program is still relocatable except that the text base should be 4KB aligned or the relocated address has the same offset with text base. Darwin want CONFIG_SYS_TEXT_BASE to be not 4KB aligned, so the relocated address should be offsetted in the same manner.
We should be careful here to use precise terms. CONFIG_SYS_TEXT_BASE is just the start address of the text segment, and if I understand correctly, it is NOT required that this is in any way aligned to a 4 KB boundary - we have to keep in mind here that the start of the text segment is not the same as the start address of the code; it can be the same, but often it is not - often other blocks of date preceed the actual program code (things like HRCW [hard reset configuration word], or strings with ID information etc., exception vector tables and/or code, etc.).
Position-independent code is usually page aligned, so gcc-aarch64 implicitly apply the restriction of text base.
Actually, u-boot for aarch64 apply another align requirement that is interrupt vector. The interrupt vector should be 2KB aligned. So, even without adrp CONFIG_SYS_TEXT_BASE should be 2KB aligned yet or relocated with the same offset, otherwise bug exist.
CONFIG_SYS_TEXT_BASE is just one number here, and it cannot be considered alone. We must also include into our considerations the linker script, because this is where actually alignments are enforced.
I think we do not need to add specific code to deal with these alignment requireents. It should be sufficient to implement these in the aarch64 linker script. This also applies for the alignment of the exception vector table.
Best regards,
Wolfgang Denk