[PATCH] MIPS: Fix __builtin_longjmp (PR 64242)

[PATCH] MIPS: Fix __builtin_longjmp (PR 64242)

[Paul Hua]

[-- Attachment #1: Type: text/plain, Size: 785 bytes --]

From 589dbe8a1c2397bfafefa4e84abe5ec6e6798928 Mon Sep 17 00:00:00 2001
From: Andrew Pinski <apinski@marvell.com>
Date: Wed, 12 Feb 2020 11:42:57 +0000
Subject: [PATCH] MIPS: Fix __builtin_longjmp (PR 64242)

The problem here is mips has its own builtin_longjmp
pattern and it was not fixed when expand_builtin_longjmp
was fixed.  We need to read the new fp and gp before
restoring the stack as the buffer might be a local
variable.

Change-Id: I3416568e260e6bde3ad5cc470fb4f2ecfa207f05
Signed-off-by: Andrew Pinski <apinski@marvell.com>

This patch from Andrew, I bootstrapped and tested on mips64el-linux-gnu.

OK for master ?

gcc/ChangeLog:

        PR middle-end/64242
        * config/mips/mips.md (builtin_longjmp): Restore the frame pointer
           and stack pointer and gp.

[-- Attachment #2: mips-fix-pr64242.patch --]
[-- Type: text/x-patch, Size: 1582 bytes --]

From 589dbe8a1c2397bfafefa4e84abe5ec6e6798928 Mon Sep 17 00:00:00 2001
From: Andrew Pinski <apinski@marvell.com>
Date: Wed, 12 Feb 2020 11:42:57 +0000
Subject: [PATCH] MIPS: Fix __builtin_longjmp (PR 64242)

The problem here is mips has its own builtin_longjmp
pattern and it was not fixed when expand_builtin_longjmp
was fixed.  We need to read the new fp and gp before
restoring the stack as the buffer might be a local
variable.

Change-Id: I3416568e260e6bde3ad5cc470fb4f2ecfa207f05
Signed-off-by: Andrew Pinski <apinski@marvell.com>
---
 gcc/config/mips/mips.md | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/gcc/config/mips/mips.md b/gcc/config/mips/mips.md
index dd9fdd1418d..8bafbe82bd3 100644
--- a/gcc/config/mips/mips.md
+++ b/gcc/config/mips/mips.md
@@ -6601,9 +6601,19 @@
 
   /* This bit is similar to expand_builtin_longjmp except that it
      restores $gp as well.  */
-  mips_emit_move (hard_frame_pointer_rtx, fp);
   mips_emit_move (pv, lab);
+  /* Restore the frame pointer and stack pointer and gp.  We must use a
+     temporary since the setjmp buffer may be a local.  */
+  fp = copy_to_reg (fp);
+  gpv = copy_to_reg (gpv);
   emit_stack_restore (SAVE_NONLOCAL, stack);
+
+  /* Ensure the frame pointer move is not optimized.  */
+  emit_insn (gen_blockage ());
+  emit_clobber (hard_frame_pointer_rtx);
+  emit_clobber (frame_pointer_rtx);
+  emit_clobber (gp);
+  mips_emit_move (hard_frame_pointer_rtx, fp);
   mips_emit_move (gp, gpv);
   emit_use (hard_frame_pointer_rtx);
   emit_use (stack_pointer_rtx);
-- 
2.17.1

Re: [PATCH] MIPS: Fix __builtin_longjmp (PR 64242)

[Paul Hua]

ping?

On Sun, Jul 12, 2020 at 2:27 PM Paul Hua <paul.hua.gm@gmail.com> wrote:
>
> From 589dbe8a1c2397bfafefa4e84abe5ec6e6798928 Mon Sep 17 00:00:00 2001
> From: Andrew Pinski <apinski@marvell.com>
> Date: Wed, 12 Feb 2020 11:42:57 +0000
> Subject: [PATCH] MIPS: Fix __builtin_longjmp (PR 64242)
>
> The problem here is mips has its own builtin_longjmp
> pattern and it was not fixed when expand_builtin_longjmp
> was fixed.  We need to read the new fp and gp before
> restoring the stack as the buffer might be a local
> variable.
>
> Change-Id: I3416568e260e6bde3ad5cc470fb4f2ecfa207f05
> Signed-off-by: Andrew Pinski <apinski@marvell.com>
>
> This patch from Andrew, I bootstrapped and tested on mips64el-linux-gnu.
>
> OK for master ?
>
> gcc/ChangeLog:
>
>         PR middle-end/64242
>         * config/mips/mips.md (builtin_longjmp): Restore the frame pointer
>            and stack pointer and gp.

Re: [PATCH] MIPS: Fix __builtin_longjmp (PR 64242)

[Jeff Law]

On Sun, 2020-07-12 at 14:27 +0800, Paul Hua wrote:
> From 589dbe8a1c2397bfafefa4e84abe5ec6e6798928 Mon Sep 17 00:00:00 2001
> From: Andrew Pinski <apinski@marvell.com>
> Date: Wed, 12 Feb 2020 11:42:57 +0000
> Subject: [PATCH] MIPS: Fix __builtin_longjmp (PR 64242)
> 
> The problem here is mips has its own builtin_longjmp
> pattern and it was not fixed when expand_builtin_longjmp
> was fixed.  We need to read the new fp and gp before
> restoring the stack as the buffer might be a local
> variable.
> 
> Change-Id: I3416568e260e6bde3ad5cc470fb4f2ecfa207f05
> Signed-off-by: Andrew Pinski <apinski@marvell.com>
> 
> This patch from Andrew, I bootstrapped and tested on mips64el-linux-gnu.
> 
> OK for master ?
> 
> gcc/ChangeLog:
> 
>         PR middle-end/64242
>         * config/mips/mips.md (builtin_longjmp): Restore the frame pointer
>            and stack pointer and gp.
OK
jeff