[ECOS] Redboot on Assabet, problem with reset_vector and UNMAPPED_PTR in vectors.S

[todd dot kallam at acterna dot com]

Hi,

I am trying to get the redboot running out of flash on my assabet/neponset
board.
I am using a snapshot of the code that I downloaded today.  My board will not
start to
boot and the problem appears to be with the reset_vector set to the wrong value.

The vectors.S file uses the following lines to set the vector value:

          vectors:
          UNMAPPED_PTR(reset_vector)                      // 0x20

The UNMAPPED_PTR macro is defined as:

          # define UNMAPPED(x) ((x)-CYGHWR_HAL_ROM_VADDR)

This expression should equate to   ((0x50000040) - 0x50000000) and equal
0x00000040,
but the map file shows the value getting set to 0x00000060.   According to the
map file,
the value should be 0x00000040.

It seems that the assembler or linker has added 0x20 to the value which also
happens
to be the offset for where vectors start.

I can manually change 0x00000060 to 0x00000040 in redboot.bin (address
0x00000020) and
the assabet boots up correctly every time.

Here is a partial listing of the objdump on the redboot.img:

------ Snip  --------------------------------------------

redboot.img:     file format elf32-littlearm

Disassembly of section .rom_vectors:

50000000 <__exception_handlers>:
50000000: e59ff018  ldr  pc, [pc, #18]   ; 50000020 <vectors>
50000004: e59ff018  ldr  pc, [pc, #18]   ; 50000024 <.undefined_instruction>
50000008: e59ff018  ldr  pc, [pc, #18]   ; 50000028 <.software_interrupt>
5000000c: e59ff018  ldr  pc, [pc, #18]   ; 5000002c <.abort_prefetch>
50000010: e59ff018  ldr  pc, [pc, #18]   ; 50000030 <.abort_data>
50000014: 00000000  andeq     r0, r0, r0
50000018: e59ff018  ldr  pc, [pc, #18]   ; 50000038 <.IRQ>
5000001c: e59ff018  ldr  pc, [pc, #18]   ; 5000003c <.FIQ>

50000020 <vectors>:
50000020: 00000060  andeq     r0, r0, r0, rrx

50000024 <.undefined_instruction>:
50000024: 50000370  andpl     r0, r0, r0, ror r3

50000028 <.software_interrupt>:
50000028: 500003f4  strpld    r0, [r0], -r4

5000002c <.abort_prefetch>:
5000002c: 5000042c  andpl     r0, r0, ip, lsr #8

50000030 <.abort_data>:
50000030: 500004a8  andpl     r0, r0, r8, lsr #9
50000034: 00000000  andeq     r0, r0, r0

50000038 <.IRQ>:
50000038: 500005b0  strplh    r0, [r0], -r0

5000003c <.FIQ>:
5000003c: 50000578  andpl     r0, r0, r8, ror r5
Disassembly of section .text:

50000040 <reset_vector>:
50000040: e3a01412  mov  r1, #301989888  ; 0x12000000
50000044: e59f26bc  ldr  r2, [pc, #6bc]  ; 50000708 <_eCos_id+0x14>
50000048: e3822a02  orr  r2, r2, #8192   ; 0x2000
5000004c: e3822901  orr  r2, r2, #16384  ; 0x4000
50000050: e5812000  str  r2, [r1]
50000054: e59f16b0  ldr  r1, [pc, #6b0]  ; 5000070c <_eCos_id+0x18>
50000058: e59f26b0  ldr  r2, [pc, #6b0]  ; 50000710 <_eCos_id+0x1c>
5000005c: e5812000  str  r2, [r1]
50000060: e59f36ac  ldr  r3, [pc, #6ac]  ; 50000714 <_eCos_id+0x20>
50000064: e3a02fbf  mov  r2, #764  ; 0x2fc
50000068: e5832000  str  r2, [r3]
5000006c: e59f26a4  ldr  r2, [pc, #6a4]  ; 50000718 <_eCos_id+0x24>
50000070: e5812000  str  r2, [r1]
50000074: e59f36a0  ldr  r3, [pc, #6a0]  ; 5000071c <_eCos_id+0x28>
50000078: e3a02064  mov  r2, #100  ; 0x64
5000007c: e5939000  ldr  r9, [r3]
50000080: e2522001  subs r2, r2, #1      ; 0x1
50000084: 1afffffc  bne  5000007c <_stext+0x3c>
---------------------------------------------------

Has anyone else experienced this problem?

I am using the 2.95.2 version of the arm-elf tools with ecos-gcc-2952.pat.

Any suggestions are welcome.

Thanks,

Todd