Commit Graph

4 Commits

Author SHA1 Message Date
Patryk Duda cab341d29c arch/common: Fix moving location counter backwards when using LLD
In GNU LD, the location counter (the 'dot' variable) always refers to
the byte offset from the start of current object as mentioned in
documentation[1]:

```
'.' actually refers to the byte offset from the start of the current
containing object. Normally this is the SECTIONS statement, whose start
address is 0, hence '.' can be used as an absolute address. If '.' is
used inside a section description however, it refers to the byte offset
from the start of that section, not an absolute address.
```

For example, if the section 'rom_start':

rom_start : {
	. = 0x400;
	_vector_start = ABSOLUTE(.);
} > FLASH

has a starting address of 0x8000000, then _vector_start will be
0x8000400

However, behavior of LLVM LLD is quite different, the value of the
location counter is always absolute (see discussion [2]), so in the
example above, the linker will return error, because it will interpret
'. = 0x400' as an attempt to move the location counter backwards.

It could be fixed by changing line to '. += 0x400' (#54796) which will
move the location counter by 0x400 for both linkers, but it would work
only when we are at the beginning of section. Consider the following
example:

rom_start : {
	. = 0x400;
	KEEP(*(.boot_hdr.conf))
	. = 0x1000;
	KEEP(*(.boot_hdr.ivt))
	KEEP(*(.boot_hdr.data))
	KEEP(*(.boot_hdr.dcd_data))
	. = 0x2000;
	_vector_start = .;
} > FLASH

In this case, _vector_start will be 0x2000, but if we change
'. = 0x2000' to '. += 0x2000', then the value of _vector_start depends
on size of data in input sections (but it's 0x3000 at least).

Actually, this example comes from final linker script when compiling
firmware for mimxrt1170_evk_cm7 board. This board failed to boot
(#55296) after #54796 was merged.

This patch introduces method compatible with both linkers. We calculate
relative offset from the beginning of the section and use that value to
calculate number of bytes by which we should move the location counter
to get CONFIG_ROM_START_OFFSET.

[1] https://sourceware.org/binutils/docs/ld/Location-Counter.html
[2] https://discourse.llvm.org/t/lld-location-counter-inside-objects

Signed-off-by: Patryk Duda <pdk@semihalf.com>
2023-07-14 12:08:02 +02:00
Declan Snyder 62d4d2a2a3 revert: "linker: rom_start_offset: add to address"
Revert commit 44628735b8

This commit broke the ability for nxp rt series to
reset except with power cycle

Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
2023-03-30 18:19:32 -04:00
Daniel Leung 44628735b8 linker: rom_start_offset: add to address instead of set
The CONFIG_ROM_START_OFFSET is supposed to be added to
the current when linking, instead of having the current
address set to it. So fix that.

Not sure why it worked up to this point, but llvm/clang/lld
complained that it could not move location counter backward.

Signed-off-by: Daniel Leung <daniel.leung@intel.com>
2023-02-22 08:46:52 -05:00
Stephanos Ioannidis 3322489d22 config: Rename TEXT_SECTION_OFFSET to ROM_START_OFFSET
The `TEXT_SECTION_OFFSET` symbol is used to specify the offset between
the beginning of the ROM area and the address of the first ROM section.

This commit renames `TEXT_SECTION_OFFSET` to `ROM_START_OFFSET` because
the first ROM section is not always the `.text` section.

Signed-off-by: Stephanos Ioannidis <root@stephanos.io>
2020-07-09 14:02:38 -04:00