RTEMS | Follow-up from "Add RISCV libdl support" (#5163)

[Chris Johns (@chris)]

Chris Johns created an issue: https://gitlab.rtems.org/rtems/rtos/rtems/-/issues/5163



The following discussion from !312 should be addressed:

- [ ] @ppisa started a [discussion](https://gitlab.rtems.org/rtems/rtos/rtems/-/merge_requests/312#note_115090 "Add RISCV libdl support"): (+5 comments)

  >>>
  I have applied your changes and tried to call `dlopen` on BeaglV-Fire from my test application

  https://github.com/ppisa/rtems-devel/tree/master/rtems-omk-template/appdl

  which calls `shell_dlopen` and tries to load simple ELF file

  build from https://github.com/ppisa/rtems-devel/blob/master/rtems-omk-template/appdl/examples/appdl_print.c

  And the result is exception

  ```
  SHLL [/] # dlopen bin/appdl_print-riscv64.o
  
  *** FATAL ***
  fatal source: 9 (RTEMS_FATAL_SOURCE_EXCEPTION)
  CPU: 0
  mstatus 0x8000000a00007880
  mcause  0x0000000000000004
  mepc    0x00000010000271a8
  ra      0x000000100002949c
  sp      0x00000010000d9ef0
  gp      0x0000001000093800
  tp      0x00000010000dabc8
  t0      0x0000000000000000
  t1      0x0000000000000001
  t2      0xffffffffffffffff
  s0      0x00000010001b97e6
  s1      0x00000010001b8660
  a0      0xffffffffffe986c2
  a1      0x0000001000089e48
  a2      0x00000010001b8940
  a3      0x0000006000000013
  a4      0x00000010000271a8
  a5      0x0000000000051ea8
  a6      0x00000010001a67a6
  a7      0x0000000000051ea8
  s2      0x0000000000051ea8
  s3      0x00000010000da008
  s4      0x00000010001a67a6
  s5      0x00000010001b88e0
  s6      0x00000010001b0ea0
  s7      0x0000000000000018
  s8      0x0000000000000060
  s9      0x0000000000000004
  s10     0x0000000000000001
  s11     0x00000010001b8d60
  t3      0x0000000000000038
  t4      0x0000000000000000
  t5      0x000000580000003c
  t6      0x0000000000000038
  fcsr    0x00000000
  RTEMS version: 6.0.0.e049757c58-modified
  RTEMS tools: 13.3.0 20240521 (RTEMS 6, RSB e52d2288c8a0df64466010839ca1eb40c6d243ed, Newlib 1b3dcfd)
  executing thread ID: 0x0a010004
  executing thread name: SHLL
  ```

  I do not have JTAG cable with special connector yet. So I have trued to decode problem from the reported mepc. GDB decodes location as

  ```
  0x10000271a8 <rtems_rtl_elf_relocate_rela+270>:      lw      a4,0(s0)
  ```

  If I try to decode to line `l * 0x10000271a8`

  then it points to the function `rtems_rtl_elf_relocate_rela`

  ```
  0x10000271a8 is in rtems_rtl_elf_relocate_rela (../../../../../git/rtems/cpukit/libdl/rtl-mdreloc-riscv.c:128).
  ```

  But if I try next location then it points to

  ```
  0x10000271aa is in rtems_rtl_elf_relocate_rela (../../../../../git/rtems/cpukit/libdl/rtl-mdreloc-riscv.c:364).
  
  ```

  ```
  359       case R_TYPE(CALL): {
  360         int64_t hi = SignExtend64(pcrel_val + 0x800, bits);
  361         write32le(where, (read32le(where) & 0xFFF) | (hi & 0xFFFFF000));
  362         int64_t hi20 = SignExtend64(pcrel_val + 0x800, bits);
  363         int64_t lo = pcrel_val - (hi20 << 12);
  364         write32le(((char *) where) + 4, (read32le(((char *) where) + 4) & 0xFFFFF) | ((lo & 0xFFF) << 20));
  365       }
  366       break;
  ```

  I have found way how to convert RTEMS generate ELF to HSS payload relatively easily by direct call of `hss-payload-generator`.

  I use `rtems-config.yaml`:

  ```
  set-name: 'PolarFire-SoC-HSS::RTEMSImage'
  hart-entry-points: {u54_1: '0x1000000000', u54_2: '0x1000000000', u54_3: '0x1000000000', u54_4: '0x1000000000'}
  payloads:
    rtems-app.elf: {exec-addr: '0x1000000000', owner-hart: u54_1, secondary-hart: u54_2, secondary-hart: u54_3, secondary-hart: u54_4, priv-mode: prv_m, skip-opensbi: true, payload-name: "rtems-app"}
  ```

  and then run in terminal

  ```
  hss-payload-generator -c rtems-config.yaml rtems-payload.bin
  ```

  >>>

-- 
View it on GitLab: https://gitlab.rtems.org/rtems/rtos/rtems/-/issues/5163
You're receiving this email because of your account on gitlab.rtems.org.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/bugs/attachments/20241122/056c2f5b/attachment-0001.htm>