[Bug target/94445] New: gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[Bug target/94445] New: gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

            Bug ID: 94445
           Summary: gcc.target/arm/cmse/cmse-15.c fails for cortex-m33
           Product: gcc
           Version: 10.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: clyon at gcc dot gnu.org
  Target Milestone: ---

I've noticed that when GCC is configured --target arm-none-abi
--with-mode=thumb --with-cpu=cortex-m33, the cmse-15.c test fails:

FAIL: gcc.target/arm/cmse/cmse-15.c   -O2   scan-assembler-times
bl\\s+__gnu_cmse_nonsecure_call 6
FAIL: gcc.target/arm/cmse/cmse-15.c   -O3 -g   scan-assembler-times
bl\\s+__gnu_cmse_nonsecure_call 6
FAIL: gcc.target/arm/cmse/cmse-15.c   -Os   scan-assembler-times
bl\\s+__gnu_cmse_nonsecure_call 6


I've looked at the -O2 case, where 8 calls to __gnu_cmse_nonsecure_call instead
of the expected 6.

The testcase can then be reduced to:
=====================================================
typedef int __attribute__ ((cmse_nonsecure_call)) ns_foo_t (void);
typedef int s_bar_t (void);

typedef int (*s_bar_ptr) (void);

int nonsecure0 (ns_foo_t * ns_foo_p)
{
  return ns_foo_p ();
}
int secure0 (s_bar_t * s_bar_p)
{
  return s_bar_p ();
}
int secure2 (s_bar_ptr s_bar_p)
{
  return s_bar_p ();
}
=====================================================

secure0 and secure2 make use of __gnu_cmse_nonsecure_call instead of doing a
normal call.

If I comment out nonsecure0(), then the test passes (only "bx r0" is generated
for secure0/secure2).


With nonsecure0() un-commented as above, I've looked at
arm_function_ok_for_sibcall() which is called 3 times.

The first time (for nonsecure0):
 <call_expr 0x7facbe3ac630
[...]
    fn <ssa_name 0x7facbe586318
        type <pointer_type 0x7facbe670888 type <function_type 0x7facbe670498
ns_foo_t>
            public unsigned SI size <integer_cst 0x7facbe57bfd8 32> unit-size
<integer_cst 0x7facbe58e000 4>
            align:32 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7facbe670930>
        visited var <parm_decl 0x7facbe39b180 ns_foo_p>
        def_stmt GIMPLE_NOP
        version:2
        ptr-info 0x7facbe39c300>


The next two times (for secure0/secure2):
 <call_expr 0x7facbe3acb10
[...]
    fn <nop_expr 0x7facbe3b34a0
        type <pointer_type 0x7facbe670888 type <function_type 0x7facbe670498
ns_foo_t>
            public unsigned SI size <integer_cst 0x7facbe57bfd8 32> unit-size
<integer_cst 0x7facbe58e000 4>
            align:32 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7facbe670930>

        arg:0 <ssa_name 0x7facbe586900 type <pointer_type 0x7facbe670b28>
            visited var <parm_decl 0x7facbe39b200 s_bar_p>
            def_stmt GIMPLE_NOP
            version:2
            ptr-info 0x7facbe3b5be8>>


 <call_expr 0x7facbe3ac060
[...]
    fn <nop_expr 0x7facbe3ba800
        type <pointer_type 0x7facbe670888 type <function_type 0x7facbe670498
ns_foo_t>
            public unsigned SI size <integer_cst 0x7facbe57bfd8 32> unit-size
<integer_cst 0x7facbe58e000 4>
            align:32 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7facbe670930>

        arg:0 <ssa_name 0x7facbe586af8 type <pointer_type 0x7facbe6707e0
s_bar_ptr>
            visited var <parm_decl 0x7facbe39b280 s_bar_p>
            def_stmt GIMPLE_NOP
            version:2
            ptr-info 0x7facbe3b8b28>>


So in the last two cases, the function being called has an incorrect 'ns_foo_t'
type.

The fact that commenting out nonsecure0 makes the test pass makes me think that
some global state is modified when handling the  cmse_nonsecure_call attribute,
but I couldn't find such a problem in arm_handle_cmse_nonsecure_call().


I haven't yet found where that function type is attached to the 'fn' node,
maybe that's where the problem is.

I've also noticed that in the first case 'fn' is an ssa_name, while in the two
other ones, it is a nop_expr.



I've been investigating this for quite some time, so any clue is appreciated.

[Bug target/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[avieira at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

avieira at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |avieira at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2020-04-02

--- Comment #1 from avieira at gcc dot gnu.org ---
Hi Christophe,

This looks to me like an issue of not building distinct types for the ns_foo_t
and s_bar_t function types.

When I first wrote this code I tested for this and it was working, so I am
wondering whether changes have been made in the way we create types in the
c-frontend.

I am trying to find out how all this works again, its been a while...

[Bug target/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[avieira at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

--- Comment #2 from avieira at gcc dot gnu.org ---
start_decl seems to be doing the right thing, investigation continues...

[Bug target/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

--- Comment #3 from Christophe Lyon <clyon at gcc dot gnu.org> ---
I also checked that arm_handle_cmse_nonsecure_call correctly duplicates the
type.

[Bug target/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[avieira at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

--- Comment #4 from avieira at gcc dot gnu.org ---
Yeah...

So far I have checked that 'gimplify_call_expr' creates the right gimple, and
up until 'gimplify_modify_expr' I can verify it does by using
gimple_call_fntype .

Though at expansion time, the 'gimple_call_fntype (stmt)' of '_5 = s_bar_p_2(D)
(); [tail call]' now has the attribute ...

So it must go wrong somewhere between gimplification and expansion, but that's
a big window and dump files won't help us :(

[Bug target/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[avieira at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

--- Comment #5 from avieira at gcc dot gnu.org ---
Yeah...

So far I have checked that 'gimplify_call_expr' creates the right gimple, and
up until 'gimplify_modify_expr' I can verify it does by using
gimple_call_fntype .

Though at expansion time, the 'gimple_call_fntype (stmt)' of '_5 = s_bar_p_2(D)
(); [tail call]' now has the attribute ...

So it must go wrong somewhere between gimplification and expansion, but that's
a big window and dump files won't help us :(

[Bug target/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[avieira at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

--- Comment #6 from avieira at gcc dot gnu.org ---
I have also identified that this only goes wrong in O2 or higher. And it
happens sometime between tailcall optimization pass 1 and 2. But there's loads
of passes in between.

[Bug ipa/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|target                      |ipa
   Target Milestone|---                         |10.0
           Assignee|clyon at gcc dot gnu.org           |marxin at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

--- Comment #7 from Martin Liška <marxin at gcc dot gnu.org> ---
It's ICF issue and I've got a patch for it.

[Bug ipa/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Martin Liska <marxin@gcc.gnu.org>:

https://gcc.gnu.org/g:55a7380213a5c16120d5c674fb42b38a3d796b57

commit r10-7530-g55a7380213a5c16120d5c674fb42b38a3d796b57
Author: Martin Liska <mliska@suse.cz>
Date:   Fri Apr 3 09:05:06 2020 +0200

    ICF: compare type attributes for gimple_call_fntypes.

            PR ipa/94445
            * ipa-icf-gimple.c (func_checker::compare_gimple_call):
              Compare type attributes for gimple_call_fntypes.

[Bug ipa/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW
      Known to fail|                            |8.4.0, 9.3.0
      Known to work|                            |10.0

--- Comment #9 from Martin Liška <marxin at gcc dot gnu.org> ---
Fixed on master, I'll backport that.

[Bug ipa/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

--- Comment #10 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-9 branch has been updated by Martin Liska
<marxin@gcc.gnu.org>:

https://gcc.gnu.org/g:69df6742cb4e15fba5ee074629a78811e61f044d

commit r9-8447-g69df6742cb4e15fba5ee074629a78811e61f044d
Author: Martin Liska <mliska@suse.cz>
Date:   Fri Apr 3 12:33:41 2020 +0200

    Backport 55a7380213a5c16120d5c674fb42b38a3d796b57

            Backport from mainline
            2020-04-03  Martin Liska  <mliska@suse.cz>

            PR ipa/94445
            * ipa-icf-gimple.c (func_checker::compare_gimple_call):
              Compare type attributes for gimple_call_fntypes.

[Bug ipa/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

--- Comment #11 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-8 branch has been updated by Martin Liska
<marxin@gcc.gnu.org>:

https://gcc.gnu.org/g:33c5b217cbc7562238f96a29393f6ca8fb97482a

commit r8-10161-g33c5b217cbc7562238f96a29393f6ca8fb97482a
Author: Martin Liska <mliska@suse.cz>
Date:   Fri Apr 3 12:36:31 2020 +0200

    Backport 55a7380213a5c16120d5c674fb42b38a3d796b57

            Backport from mainline
            2020-04-03  Martin Liska  <mliska@suse.cz>

            PR ipa/94445
            * ipa-icf-gimple.c (func_checker::compare_gimple_call):
              Compare type attributes for gimple_call_fntypes.

[Bug ipa/94445] gcc.target/arm/cmse/cmse-15.c fails for cortex-m33

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94445

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED
      Known to fail|8.4.0, 9.3.0                |
      Known to work|                            |8.4.1, 9.3.1

--- Comment #12 from Martin Liška <marxin at gcc dot gnu.org> ---
Fixed on active branches.