Differences

This shows you the differences between two versions of the page.

--- wswan:guide:elf_quirks [2024/09/22 09:16] – [Address space] asie
+++ wswan:guide:elf_quirks [2024/09/22 11:58] (current) – asie
@@ Line 1: / Line 1: @@
-====== WonderSwan ELF quirks ======
+====== wswan toolchain ELF quirks ======
 The ELF format created by Wonderful has some quirks, owing both to the non-upstream binutils fork used and to the specifics of the ROM format itself.
@@ Line 5: / Line 5: @@
 In general, gcc-ia16 ELF files follow the 32-bit x86 ELF format.
-===== Address space =====
+===== Segment relocations =====
-ELF addresses ''0x80000000 - 0xFFFFFFFF'' correspond to ROM addresses.
+binutils-ia16 introduces a custom set of relocations to support segmentation - the segelf specification as proposed by H. Peter Anvin. In particular, ELFs emitted by Wonderful currently use the following relocations:
+  * ''R_386_32'' (in DWARF debug sections; uses 32-bit ELF addresses above)
+  * ''R_386_16''
+  * ''R_386_PC16''
+  * ''R_386_SEG16'' (segelf)
+  * ''R_386_SUB16'' (segelf)
+In addition, ''R_386_SUB32'' relocations are improperly emitted in DWARF debug sections by some versions of the toolchain; these should be ignored.
+More extensive documentation for these relocations is available [[https://codeberg.org/tkchia/build-ia16/src/branch/master/elf16-writeup.md|here]].
+===== Address space (wswan) =====
+ELF addresses ''0x80000000 - 0xFFFFFFFF'' correspond to addresses in ROM.
 <code>
@@ Line 15: / Line 29: @@
   ||| |||| |||| |||| |||| |||| |||| ||||
   ||| |||| |||| ++++-++++-++++-++++-++++- CPU address (bits 0-19)
-  +++-++++-++++-------------------------- ROM bank (bits 30-20)
+  +++-++++-++++-------------------------- 64KB ROM bank (bits 30-20)
                                           Corresponds to banks 000 - 7FF (128 MB maximum)
                                           The last ROM bank uses addresses FFFxxxxx
 </code>
+To turn this into a physical address in the cartridge ROM's space, you should use the following equation: ''%%((addr & 0xFFFF) | ((addr & 0x7FF00000) >> 4)) & rom_mask%%''.
+The high four bits of the CPU address are ignored for physical ROM address transformations. This is because their purpose is to tell the compiler/linker at what address the CPU accesses the 64KB bank, not where it is physically stored on the cartridge. For example:
+  * The ELF address ''FFFF0000'' means ''ROM bank 7FF, offset 0000, accessed at F0000 by the CPU'',
+  * The ELF address ''FFF30000'' means ''ROM bank 7FF, offset 0000, accessed at 30000 by the CPU''.
 ELF addresses ''0x00000000 - 0x0000FFFF'' correspond to IRAM addresses.
@@ Line 30: / Line 51: @@
 </code>
-ELF addresses ''0x00010000 - 0x0001FFFF'' currently correspond to SRAM addresses, but this may be changed in a future version (to support >64KB SRAM), so please don't rely on it.
+ELF addresses ''0x00n10000 - 0x00n1FFFF'' currently correspond to SRAM addresses (where ''n'' is the SRAM bank), but this may be changed in a future version, so please don't rely on it.
+===== Address space (wwitch) =====
+All of this may be changed in a future version, so please don't rely on it.
+ELF VMA addresses ''0x00000000 - 0x0000FFFF'' correspond to the 64K of code placed in NOR flash.
+ELF VMA addresses ''0x00010000 - 0x0001FFFF'' correspond to the 64K of data placed in SRAM.