[llvm-bugs] [Bug 36031] New: [arm] After r203575, Bad machine code: Using an undefined physical register

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=36031

            Bug ID: 36031
           Summary: [arm] After r203575, Bad machine code: Using an
                    undefined physical register
           Product: new-bugs
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: dimitry at andric.com
                CC: llvm-bugs at lists.llvm.org

As reported in https://bugs.freebsd.org/225278, building the FreeBSD
mail/deforaos-mailer port for armv6 results in "Bad machine code: Using an
undefined physical register" errors.

These errors still occur with trunk r322755.

Minimized test case:

/* clang -cc1 -triple armv6kz -S -target-cpu arm1176jzf-s -O1 -stack-protector
3 message-minimized.c */
static struct {
  const int a;
  int (*b)();
} c[];
int d;
void e() { c[d].b(0, 0); }

Error output:

# Machine code for function e: NoPHIs, TracksLiveness
Frame Objects:
  fi#0: size=4, align=4, at location [SP]

%bb.0: derived from LLVM BB %entry
        %1:gpr = LDRLIT_ga_pcrel_ldr target-flags(arm-got) @d; mem:LD4[GOT]
        %2:gpr = LDRi12 %1, 0, 14, %noreg;
mem:LD4[@d](tbaa=!4)(dereferenceable)
        %3:gpr = LDRLIT_ga_pcrel_ldr target-flags(arm-got) @__stack_chk_guard;
mem:LD4[GOT]
        %4:gpr = LDRi12 %3, 0, 14, %noreg; mem:Volatile LD4[@__stack_chk_guard]
        STRi12 %4, %stack.0.StackGuardSlot, 0, 14, %noreg; mem:Volatile
ST4[FixedStack0]
        %9:gpr = LDRi12 %3, 0, 14, %noreg; mem:Volatile LD4[@__stack_chk_guard]
        %10:gpr = LDRi12 %stack.0.StackGuardSlot, 0, 14, %noreg; mem:Volatile
LD4[FixedStack0]
        dead %11:gpr = SUBrr %9, %10, 14, %noreg, def %cpsr
        Bcc %bb.1, 1, killed %cpsr
        B %bb.2
    Successors according to CFG: %bb.2(0x7ffff800 / 0x80000000 = 100.00%)
%bb.1(0x00000800 / 0x80000000 = 0.00%)

%bb.1: derived from LLVM BB %entry
    Predecessors according to CFG: %bb.0
        ADJCALLSTACKDOWN 0, 0, 14, %noreg, implicit-def dead %sp, implicit %sp
        BL &__stack_chk_fail, <regmask %lr %d8 %d9 %d10 %d11 %d12 %d13 %d14
%d15 %q4 %q5 %q6 %q7 %r4 %r5 %r6 %r7 %r8 %r9 %r10 %r11 %s16 %s17 %s18 %s19 %s20
%s21 %s22 %s23 %s24 %s25 %s26 %s27 and 35 more...>, implicit-def dead %lr,
implicit %sp, implicit-def %sp
        ADJCALLSTACKUP 0, 0, 14, %noreg, implicit-def dead %sp, implicit %sp

%bb.2: derived from LLVM BB %entry
    Predecessors according to CFG: %bb.0
        %0:gpr = LDRLIT_ga_pcrel @c
        %5:gpr = ADDrsi %0, %2, 26, 14, %noreg, %noreg
        %7:tcgpr = LDRi12 %5, 4, 14, %noreg; mem:LD4[%2](tbaa=!8)
        TCRETURNri %7, implicit %sp, implicit %r0, implicit killed %r1

# End machine code for function e.

*** Bad machine code: Using an undefined physical register ***
- function:    e
- basic block: %bb.2 entry (0x8051cc618)
- instruction: TCRETURNri
- operand 2:   implicit %r0

*** Bad machine code: Using an undefined physical register ***
- function:    e
- basic block: %bb.2 entry (0x8051cc618)
- instruction: TCRETURNri
- operand 3:   implicit killed %r1
fatal error: error in backend: Found 2 machine code errors.

It seems to have something to do with stack protector, since using
-stack-protector 2 makes the problem go away.

Bisection showed this regressed with https://reviews.llvm.org/rL203575 ("ARM:
enable tail call optimisation on Thumb 2") by Saleem Abdulrasool.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20180121/c3faa733/attachment.html>