[Bug gas/23956] RISC-V 4-operand add doesn't check for %tprel_add

[wilson at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=23956

Jim Wilson <wilson at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2018-12-07
           Assignee|unassigned at sourceware dot org   |wilson at gcc dot 
gnu.org
     Ever confirmed|0                           |1

--- Comment #1 from Jim Wilson <wilson at gcc dot gnu.org> ---
I consider it a bug, but not serious enough that I had gotten around to trying
to fix it yet.  The support for the 4-operand add is overloaded with the amo*
instruction support that requires addresses with a constant 0 immediate offset.
 It is an odd design.  I don't know why it was done this way.  The fact that we
accept 0 for an amo* instruction is why zero gets accepted for the 4-operand
add, which looked like a fairly harmless bug to me.  But looking at this
closer, I see that the same overload means that we can use %tprel_add in an
amo* instruction, which is a more serious problem that I think justifies a fix
now.

rohan:2097$ cat tmp2.s
        amoadd.w x8,x9,%tprel_add(i)(x10)
        .globl  i
        .section        .tbss,"awT",@nobits
        .align  2
        .type   i, @object
        .size   i, 4
i:
        .zero   4
rohan:2098$ ./as-new tmp2.s
rohan:2099$ ../binutils/objdump -dr a.out

a.out:     file format elf32-littleriscv


Disassembly of section .text:

00000000 <.text>:
   0:   0095242f                amoadd.w        s0,s1,(a0)
                        0: R_RISCV_TPREL_ADD    i
                        0: R_RISCV_RELAX        *ABS*
rohan:2100$

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