[Bug ld/32367] New: Cannot b.w to thumb symbol from symbol

[maxmsonderegger at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=32367

            Bug ID: 32367
           Summary: Cannot b.w to thumb symbol from symbol
           Product: binutils
           Version: 2.43.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: ld
          Assignee: unassigned at sourceware dot org
          Reporter: maxmsonderegger at gmail dot com
  Target Milestone: ---

Running the below command generates a stub to transition from thumb to arm,
when branching to a provided symbol. The symbol is not marked as arm
explicitly, so I expect the compiler and linker to assume it is thumb and emit
a b.w to the target. This prevents linking multiple large thumb object files
together and I do not believe it is intended behavior. I am happy to submit a
bug fix, but am new to contributing to binutils and would like to know what you
all believe is the best way to address this? Assembly directives, linker
patching, linker script hints?

Things I tried:
Setting . = 0x40001 has no effect. Using .thumb_set, .set, etc. on
___top_of_text to provide hints to the assembler has no effect.  Compiling with
-mthumb has no effect. Setting the arch to one of the thumb only architectures
(e.g. armv7-m) does make it do the expected behavior and emit a b.w to the
location, however this feels like a kludge and limits my target architectures.

script.ld:
SECTIONS {
        . = 0x40000;
        PROVIDE( ___top_of_text = .);
        .text : {
                *(.text)
        }
}

---

test.s:
.syntax unified
.thumb
mov r1, r2
mov r3, r4
b.w ___top_of_text

---

%> arm-none-eabi-as -o test.o test.s && arm-none-eabi-ld -o test test.o -T
script.ld && arm-none-eabi-objdump -d test.o && arm-none-eabi-objdump -x test.o
&& arm-none-eabi-objdump -d test

test.o:     file format elf32-littlearm


Disassembly of section .text:

00000000 <.text>:
   0:   4611            mov     r1, r2
   2:   4623            mov     r3, r4
   4:   f7ff bffe       b.w     0 <___top_of_text>

test.o:     file format elf32-littlearm
test.o
architecture: armv6t2, flags 0x00000011:
HAS_RELOC, HAS_SYMS
start address 0x00000000
private flags = 0x5000000: [Version5 EABI]

Sections:
Idx Name          Size      VMA       LMA       File off  Algn
  0 .text         00000008  00000000  00000000  00000034  2**1
                  CONTENTS, ALLOC, LOAD, RELOC, READONLY, CODE
  1 .data         00000000  00000000  00000000  0000003c  2**0
                  CONTENTS, ALLOC, LOAD, DATA
  2 .bss          00000000  00000000  00000000  0000003c  2**0
                  ALLOC
  3 .ARM.attributes 00000014  00000000  00000000  0000003c  2**0
                  CONTENTS, READONLY
SYMBOL TABLE:
00000000 l    d  .text  00000000 .text
00000000 l    d  .data  00000000 .data
00000000 l    d  .bss   00000000 .bss
00000000 l    d  .ARM.attributes        00000000 .ARM.attributes
00000000         *UND*  00000000 ___top_of_text


RELOCATION RECORDS FOR [.text]:
OFFSET   TYPE              VALUE
00000004 R_ARM_THM_JUMP24  ___top_of_text



test:     file format elf32-littlearm


Disassembly of section .text:

00040000 <___top_of_text>:
   40000:       4611            mov     r1, r2
   40002:       4623            mov     r3, r4
   40004:       f000 b800       b.w     40008 <_____top_of_text_from_thumb>

00040008 <_____top_of_text_from_thumb>:
   40008:       4778            bx      pc
   4000a:       e7fd            b.n     40008 <_____top_of_text_from_thumb>
   4000c:       eafffffb        b       40000 <___top_of_text>

-- 
You are receiving this mail because:
You are on the CC list for the bug.