From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugzilla@gcc.gnu.org>
Received: by sourceware.org (Postfix, from userid 48)
 id 09FF2385801D; Sun,  8 Nov 2020 00:53:35 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 09FF2385801D
From: "martingalvan at sourceware dot org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug libgcc/97754] New: arm/lib1funcs.S (RETLDM): CFI may be incorrect
Date: Sun, 08 Nov 2020 00:53:34 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: new
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: libgcc
X-Bugzilla-Version: 11.0
X-Bugzilla-Keywords: 
X-Bugzilla-Severity: normal
X-Bugzilla-Who: martingalvan at sourceware dot org
X-Bugzilla-Status: UNCONFIRMED
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: bug_id short_desc product version bug_status
 bug_severity priority component assigned_to reporter target_milestone
Message-ID: <bug-97754-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-BeenThere: gcc-bugs@gcc.gnu.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Gcc-bugs mailing list <gcc-bugs.gcc.gnu.org>
List-Unsubscribe: <https://gcc.gnu.org/mailman/options/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=unsubscribe>
List-Archive: <https://gcc.gnu.org/pipermail/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-request@gcc.gnu.org?subject=help>
List-Subscribe: <https://gcc.gnu.org/mailman/listinfo/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=subscribe>
X-List-Received-Date: Sun, 08 Nov 2020 00:53:35 -0000

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D97754

            Bug ID: 97754
           Summary: arm/lib1funcs.S (RETLDM): CFI may be incorrect
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libgcc
          Assignee: unassigned at gcc dot gnu.org
          Reporter: martingalvan at sourceware dot org
  Target Milestone: ---

Hi all, I'm looking at the RETLDM macro for the ARM libgcc code:

.macro  RETLDM  regs=3D, cond=3D, unwind=3D, dirn=3Dia
#if defined (__INTERWORKING__)
        .ifc "\regs",""
        ldr\cond        lr, [sp], #8
        .else
# if defined(__thumb2__)
        pop\cond        {\regs, lr}
# else
        ldm\cond\dirn   sp!, {\regs, lr}
# endif
        .endif
        .ifnc "\unwind", ""
        /* Mark LR as restored.  */
97:     cfi_pop 97b - \unwind, 0xe, 0x0
        .endif
        bx\cond lr
#else
<...>
.endm

Here we can see that for the __INTERWORKING__ case, some CFI data is emitte=
d if
we have an 'unwind' argument. However, I believe this may be wrong if a RET=
LDM
macro appears in the middle of another function. For the __INTERWORKING__ c=
ase,
the resulting output would be e.g:

pop {r4, r5, lr}
cfi_pop <...>
bx  lr
<rest of the function>

Here, 'cfi_pop' would affect the rest of the function unless the caller is
careful enough to restore the CFI status for lr after doing RETLDM. This can
happen e.g. if we pass a 'cond' argument, which means the branch may not be
taken.

Unless I'm mistaken, the way to fix this would be to introduce a couple of
.cfi_remember/restore_state directives:

    .ifnc "\unwind", ""
        /* Mark LR as restored.  */
        .cfi_remember_state
97:     cfi_pop 97b - \unwind, 0xe, 0x0
    .endif

    bx\cond lr

    .ifnc "\unwind", ""
        .cfi_restore_state
    .endif

If this is correct, I'll send in a patch to add this change.

Speaking of this file, I have a couple more questions:

1. The file is defining macros like cfi_start, cfi_push, etc which insert t=
he
CFI data directly in binary form. Why can't we use gas' .cfi_* directives a=
s we
do elsewhere? If we can, I'll send a separate patch to add this change as w=
ell.
2. Why is the __INTERWORKING__ macro only defined for ARMv4t? I'm not saying
that this is wrong, just curious.

Thanks.=