[Bug ld/27566] [RISC-V] relocation truncated to fit: R_RISCV_GPREL_I against aymbol

[address@hidden]

https://sourceware.org/bugzilla/show_bug.cgi?id=27566

--- Comment #3 from lifang_xia@c-sky.com ---
Hi Nelson,

I agree with you, the previous patch is not a best choice.

I have do some debug today.
I add a print here:

===========================================
diff --git a/bfd/elfnn-riscv.c b/bfd/elfnn-riscv.c
index 364d67b..e774cff 100644
--- a/bfd/elfnn-riscv.c
+++ b/bfd/elfnn-riscv.c
@@ -4247,14 +4247,23 @@ _bfd_riscv_relax_lui (bfd *abfd,
        max_alignment = (bfd_vma) 1 <<
sym_sec->output_section->alignment_power;
     }

+  printf ("gp: %lx, sym val: %lx\n", gp, symval);
   /* Is the reference in range of x0 or gp?
      Valid gp range conservatively because of alignment issue.  */
   if (undefined_weak
       || (VALID_ITYPE_IMM (symval)
          || (symval >= gp
==============================================

And I got a result:
-----------------------------------------------
xialf@magics:build-linux$ ./ld/ld-new -o a a.o -e main  -z norelro
gp: 11d90, sym val: 11c10
gp: 11d90, sym val: 11c10
xialf@magics:build-linux$ readelf -s a |grep global
     5: 0000000000011d90     0 NOTYPE  GLOBAL DEFAULT  ABS __global_pointer$
-----------------------------------------------

>From this, GP is not changed, that means DATA_SEGMENT_ALIGN  has been finished
while relaxing.Therefore we can just consider about COMMON_PAGESIZE, not
MAX_PAGESIZE+COMMON_PAGESIZE.

=====================================================
For -z relro

xialf@magics:build-linux$ ./ld/ld-new -o a a.o -e main  -z relro
gp: 12800, sym val: 11c10
gp: 12800, sym val: 11c10
gp: 11d90, sym val: 11c10
gp: 11d90, sym val: 11c10
xialf@magics:build-linux$ readelf -s a |grep global
     5: 0000000000012800     0 NOTYPE  GLOBAL DEFAULT  ABS __global_pointer$
=======================================================
DATA_SEGMENT_RELRO_END has been aligned to 0x12000, and gp is 0x12800.
=======================================================
common-page-size=0x4000

xialf@magics:build-linux$ ./ld/ld-new -o a a.o -e main -z
common-page-size=0x4000 -z relro
gp: 14800, sym val: 11c10
gp: 14800, sym val: 11c10
gp: 11d90, sym val: 11c10
gp: 11d90, sym val: 11c10
xialf@magics:build-linux$ readelf -s a |grep global
     5: 0000000000014800     0 NOTYPE  GLOBAL DEFAULT  ABS __global_pointer$
========================================================
DATA_SEGMENT_RELRO_END has been aligned to 0x14000, gp is 0x14800.

It is not check the imm with (MAX_PAGESIZE + COMMON_PAGE_SIZE) or
(COMMON_PAGESIZE); It need to align to COMMON_PAGE_SIZE.

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