diff mbox series

[V2] gas: gcfg: fix handling of non-local direct jmps in gcfg

Message ID 20240328075909.4134867-1-indu.bhagat@oracle.com
State New
Headers
Series [V2] gas: gcfg: fix handling of non-local direct jmps in gcfg |

Checks

Context Check Description
linaro-tcwg-bot/tcwg_binutils_build--master-arm success Testing passed
linaro-tcwg-bot/tcwg_binutils_build--master-aarch64 success Testing passed
linaro-tcwg-bot/tcwg_binutils_check--master-aarch64 success Testing passed
linaro-tcwg-bot/tcwg_binutils_check--master-arm success Testing passed

Commit Message

Indu Bhagat March 28, 2024, 7:59 a.m. UTC 
  [Changes in V2]
  - Add function level comments for ginsn_direct_local_jump_p.
  - Add a jmp to locally defined label in the testcase to make it more
    complete.  Also, add a jmp to label defined in another function.
    The testcase now generates an additional warning (as expected):
    "Warning: GINSN: found unreachable code in func 'bar'"
  - Use formatting style consistent with other testcases.
[End of changes in V2]

The ginsn infrastructure in GAS includes the ability to create a GCFG
(ginsn CFG).  A GCFG is currently used for SCFI passes.

This patch fixes the following invalid assumptions / code blocks:
 - The function ginsn_direct_local_jump_p () was erroneously _not_
   checking whether the symbol is locally defined (i.e., within the
   scope of the code block for which GCFG is desired).  Fix the code
   to do so.
 - Similarly, the GCFG creation code, in gcfg_build () itself had an
   assumption that a GINSN_TYPE_JUMP to a non-local symbol will not be
   seen.  The latter can indeed be seen, and in fact, needs to be treated
   the same way as an exit from the function in terms of control-flow.

gas/
        * ginsn.c (ginsn_direct_local_jump_p): Check if the symbol
	is local to the code block or function being assembled.
        (add_bb_at_ginsn): Remove buggy assumption.
        (frch_ginsn_data_append): Direct jmps do not disqualify a stream
	of ginsns from GCFG creation.

gas/testsuite/
	* gas/scfi/x86_64/scfi-cfg-3.d: New test.
	* gas/scfi/x86_64/scfi-cfg-3.l: New test.
	* gas/scfi/x86_64/scfi-cfg-3.s: New test.
	* gas/scfi/x86_64/scfi-x86-64.exp: Add new test.

gas: gcfg: address review comments for local jump issue

ChangeLog:
	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d
	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l
	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s
---
 gas/ginsn.c                                   | 47 +++++++++++--------
 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d    | 32 +++++++++++++
 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l    |  3 ++
 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s    | 43 +++++++++++++++++
 gas/testsuite/gas/scfi/x86_64/scfi-x86-64.exp |  2 +
 5 files changed, 108 insertions(+), 19 deletions(-)
 create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d
 create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l
 create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s

Comments

Jan Beulich March 28, 2024, 10:10 a.m. UTC | #1 
On 28.03.2024 08:59, Indu Bhagat wrote:
> [Changes in V2]
>   - Add function level comments for ginsn_direct_local_jump_p.
>   - Add a jmp to locally defined label in the testcase to make it more
>     complete.  Also, add a jmp to label defined in another function.
>     The testcase now generates an additional warning (as expected):
>     "Warning: GINSN: found unreachable code in func 'bar'"

For this - can't you simply omit the unreachable pair of insns?

Other than that, looks plausible, so feel free to put in.

>   - Use formatting style consistent with other testcases.
> [End of changes in V2]

As a minor remark, this entire section would better go ...

> The ginsn infrastructure in GAS includes the ability to create a GCFG
> (ginsn CFG).  A GCFG is currently used for SCFI passes.
> 
> This patch fixes the following invalid assumptions / code blocks:
>  - The function ginsn_direct_local_jump_p () was erroneously _not_
>    checking whether the symbol is locally defined (i.e., within the
>    scope of the code block for which GCFG is desired).  Fix the code
>    to do so.
>  - Similarly, the GCFG creation code, in gcfg_build () itself had an
>    assumption that a GINSN_TYPE_JUMP to a non-local symbol will not be
>    seen.  The latter can indeed be seen, and in fact, needs to be treated
>    the same way as an exit from the function in terms of control-flow.
> 
> gas/
>         * ginsn.c (ginsn_direct_local_jump_p): Check if the symbol
> 	is local to the code block or function being assembled.
>         (add_bb_at_ginsn): Remove buggy assumption.
>         (frch_ginsn_data_append): Direct jmps do not disqualify a stream
> 	of ginsns from GCFG creation.
> 
> gas/testsuite/
> 	* gas/scfi/x86_64/scfi-cfg-3.d: New test.
> 	* gas/scfi/x86_64/scfi-cfg-3.l: New test.
> 	* gas/scfi/x86_64/scfi-cfg-3.s: New test.
> 	* gas/scfi/x86_64/scfi-x86-64.exp: Add new test.
> 
> gas: gcfg: address review comments for local jump issue
> 
> ChangeLog:
> 	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d
> 	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l
> 	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s
> ---
>  gas/ginsn.c                                   | 47 +++++++++++--------
>  gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d    | 32 +++++++++++++
>  gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l    |  3 ++
>  gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s    | 43 +++++++++++++++++
>  gas/testsuite/gas/scfi/x86_64/scfi-x86-64.exp |  2 +
>  5 files changed, 108 insertions(+), 19 deletions(-)
>  create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d
>  create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l
>  create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s

... below here, perhaps with another --- separator. This way it's clear
that the revlog isn't going to be part of the commit message.

Jan
Indu Bhagat March 28, 2024, 4:26 p.m. UTC | #2 
On 3/28/24 03:10, Jan Beulich wrote:
> On 28.03.2024 08:59, Indu Bhagat wrote:
>> [Changes in V2]
>>    - Add function level comments for ginsn_direct_local_jump_p.
>>    - Add a jmp to locally defined label in the testcase to make it more
>>      complete.  Also, add a jmp to label defined in another function.
>>      The testcase now generates an additional warning (as expected):
>>      "Warning: GINSN: found unreachable code in func 'bar'"
> 
> For this - can't you simply omit the unreachable pair of insns?
> 

We do omit those instructions from GCFG and SCFI.  GAS does, however, 
warn for now. Do you mean skip this warning too?

> Other than that, looks plausible, so feel free to put in.
> 
>>    - Use formatting style consistent with other testcases.
>> [End of changes in V2]
> 
> As a minor remark, this entire section would better go ...
> 

Yes, I usually get rid of this before committing.  But I agree, its 
better put below after the separator.

>> The ginsn infrastructure in GAS includes the ability to create a GCFG
>> (ginsn CFG).  A GCFG is currently used for SCFI passes.
>>
>> This patch fixes the following invalid assumptions / code blocks:
>>   - The function ginsn_direct_local_jump_p () was erroneously _not_
>>     checking whether the symbol is locally defined (i.e., within the
>>     scope of the code block for which GCFG is desired).  Fix the code
>>     to do so.
>>   - Similarly, the GCFG creation code, in gcfg_build () itself had an
>>     assumption that a GINSN_TYPE_JUMP to a non-local symbol will not be
>>     seen.  The latter can indeed be seen, and in fact, needs to be treated
>>     the same way as an exit from the function in terms of control-flow.
>>
>> gas/
>>          * ginsn.c (ginsn_direct_local_jump_p): Check if the symbol
>> 	is local to the code block or function being assembled.
>>          (add_bb_at_ginsn): Remove buggy assumption.
>>          (frch_ginsn_data_append): Direct jmps do not disqualify a stream
>> 	of ginsns from GCFG creation.
>>
>> gas/testsuite/
>> 	* gas/scfi/x86_64/scfi-cfg-3.d: New test.
>> 	* gas/scfi/x86_64/scfi-cfg-3.l: New test.
>> 	* gas/scfi/x86_64/scfi-cfg-3.s: New test.
>> 	* gas/scfi/x86_64/scfi-x86-64.exp: Add new test.
>>
>> gas: gcfg: address review comments for local jump issue
>>
>> ChangeLog:
>> 	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d
>> 	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l
>> 	* gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s
>> ---
>>   gas/ginsn.c                                   | 47 +++++++++++--------
>>   gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d    | 32 +++++++++++++
>>   gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l    |  3 ++
>>   gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s    | 43 +++++++++++++++++
>>   gas/testsuite/gas/scfi/x86_64/scfi-x86-64.exp |  2 +
>>   5 files changed, 108 insertions(+), 19 deletions(-)
>>   create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d
>>   create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l
>>   create mode 100644 gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s
> 
> ... below here, perhaps with another --- separator. This way it's clear
> that the revlog isn't going to be part of the commit message.
> 
> Jan
Jan Beulich March 28, 2024, 4:35 p.m. UTC | #3 
On 28.03.2024 17:26, Indu Bhagat wrote:
> On 3/28/24 03:10, Jan Beulich wrote:
>> On 28.03.2024 08:59, Indu Bhagat wrote:
>>> [Changes in V2]
>>>    - Add function level comments for ginsn_direct_local_jump_p.
>>>    - Add a jmp to locally defined label in the testcase to make it more
>>>      complete.  Also, add a jmp to label defined in another function.
>>>      The testcase now generates an additional warning (as expected):
>>>      "Warning: GINSN: found unreachable code in func 'bar'"
>>
>> For this - can't you simply omit the unreachable pair of insns?
> 
> We do omit those instructions from GCFG and SCFI.  GAS does, however, 
> warn for now. Do you mean skip this warning too?

No, I'm talking of only this one new testcase.

Jan
Indu Bhagat March 28, 2024, 11:03 p.m. UTC | #4 
On 3/28/24 09:35, Jan Beulich wrote:
> On 28.03.2024 17:26, Indu Bhagat wrote:
>> On 3/28/24 03:10, Jan Beulich wrote:
>>> On 28.03.2024 08:59, Indu Bhagat wrote:
>>>> [Changes in V2]
>>>>     - Add function level comments for ginsn_direct_local_jump_p.
>>>>     - Add a jmp to locally defined label in the testcase to make it more
>>>>       complete.  Also, add a jmp to label defined in another function.
>>>>       The testcase now generates an additional warning (as expected):
>>>>       "Warning: GINSN: found unreachable code in func 'bar'"
>>> For this - can't you simply omit the unreachable pair of insns?
>> We do omit those instructions from GCFG and SCFI.  GAS does, however,
>> warn for now. Do you mean skip this warning too?
> No, I'm talking of only this one new testcase.

OK. I removed the unreachable pair of insns and the warning from the 
testcase, and pushed.

Thanks
diff mbox series

Patch

diff --git a/gas/ginsn.c b/gas/ginsn.c
index 661f51d23c5..492e161876b 100644
--- a/gas/ginsn.c
+++ b/gas/ginsn.c
@@ -444,16 +444,26 @@  ginsn_indirect_jump_p (ginsnS *ginsn)
   return ret_p;
 }
 
+/* Return whether the GINSN is an unconditional jump to a label which is
+   defined locally in the scope of the block of insns, which are currently
+   being processed for GCFG creation.  */
+
 static bool
 ginsn_direct_local_jump_p (ginsnS *ginsn)
 {
-  bool ret_p = false;
+  bool local_p = false;
+  const symbolS *taken_label;
+
   if (!ginsn)
-    return ret_p;
+    return local_p;
 
-  ret_p |= (ginsn->type == GINSN_TYPE_JUMP
-	    && ginsn->src[0].type == GINSN_SRC_SYMBOL);
-  return ret_p;
+  if (ginsn->type == GINSN_TYPE_JUMP
+      && ginsn->src[0].type == GINSN_SRC_SYMBOL)
+    {
+      taken_label = ginsn->src[0].sym;
+      local_p = (label_ginsn_map_find (taken_label) != NULL);
+    }
+  return local_p;
 }
 
 static char *
@@ -785,16 +795,13 @@  add_bb_at_ginsn (const symbolS *func, gcfgS *gcfg, ginsnS *ginsn, gbbS *prev_bb,
 	  || ginsn->type == GINSN_TYPE_JUMP_COND
 	  || ginsn->type == GINSN_TYPE_RETURN)
 	{
-	  /* Indirect Jumps or direct jumps to symbols non-local to the
-	     function must not be seen here.  The caller must have already
-	     checked for that.  */
+	  /* Indirect jumps must not be seen here.  The caller must have
+	     already checked for that.  */
 	  gas_assert (!ginsn_indirect_jump_p (ginsn));
-	  if (ginsn->type == GINSN_TYPE_JUMP)
-	    gas_assert (ginsn_direct_local_jump_p (ginsn));
 
-	  /* Direct Jumps.  May include conditional or unconditional change of
-	     flow.  What is important for CFG creation is that the target be
-	     local to function.  */
+	  /* Handle direct jumps.  For unconditional direct jumps, where the
+	     target is not local to the function, treat them later as similar
+	     to an exit from function (in the else block).  */
 	  if (ginsn->type == GINSN_TYPE_JUMP_COND
 	      || ginsn_direct_local_jump_p (ginsn))
 	    {
@@ -822,10 +829,14 @@  add_bb_at_ginsn (const symbolS *func, gcfgS *gcfg, ginsnS *ginsn, gbbS *prev_bb,
 	      /* Add the bb for the fall through path.  */
 	      find_or_make_bb (func, gcfg, ginsn->next, prev_bb, errp);
 	    }
-	 else if (ginsn->type == GINSN_TYPE_RETURN)
+	 else
 	   {
-	     /* We'll come back to the ginsns following GINSN_TYPE_RETURN
-		from another path if they are indeed reachable code.  */
+	     gas_assert (ginsn->type == GINSN_TYPE_RETURN
+			 || (ginsn->type == GINSN_TYPE_JUMP
+			     && !ginsn_direct_local_jump_p (ginsn)));
+	     /* We'll come back to the ginsns following GINSN_TYPE_RETURN or
+		other (non-local) unconditional jmps from another path if they
+		are indeed reachable code.  */
 	     break;
 	   }
 
@@ -1083,9 +1094,7 @@  frch_ginsn_data_append (ginsnS *ginsn)
     {
       temp->id = ++id;
 
-      if (ginsn_indirect_jump_p (temp)
-	  || (ginsn->type == GINSN_TYPE_JUMP
-	      && !ginsn_direct_local_jump_p (temp)))
+      if (ginsn_indirect_jump_p (temp))
 	frchain_now->frch_ginsn_data->gcfg_apt_p = false;
 
       if (listing & LISTING_GINSN_SCFI)
diff --git a/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d b/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d
new file mode 100644
index 00000000000..dfc2782d1b9
--- /dev/null
+++ b/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.d
@@ -0,0 +1,32 @@ 
+#as: --scfi=experimental -W
+#as:
+#objdump: -Wf
+#name: Synthesize CFI in presence of control flow 3
+#...
+Contents of the .eh_frame section:
+
+00000000 0+0014 0+0000 CIE
+  Version:               1
+  Augmentation:          "zR"
+  Code alignment factor: 1
+  Data alignment factor: -8
+  Return address column: 16
+  Augmentation data:     1b
+  DW_CFA_def_cfa: r7 \(rsp\) ofs 8
+  DW_CFA_offset: r16 \(rip\) at cfa-8
+  DW_CFA_nop
+  DW_CFA_nop
+
+0+0018 0+0010 0+001c FDE cie=00000000 pc=0+0000..0+001e
+  DW_CFA_nop
+#...
+
+0+002c 0+0018 0+0030 FDE cie=00000000 pc=0+001e..0+002b
+  DW_CFA_advance_loc: 1 to 0+001f
+  DW_CFA_def_cfa_offset: 16
+  DW_CFA_offset: r6 \(rbp\) at cfa-16
+  DW_CFA_advance_loc: 3 to 0+0022
+  DW_CFA_def_cfa_register: r6 \(rbp\)
+  DW_CFA_nop
+
+#pass
diff --git a/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l b/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l
new file mode 100644
index 00000000000..eda60a41fef
--- /dev/null
+++ b/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.l
@@ -0,0 +1,3 @@ 
+.*Assembler messages:
+.*9: Warning: SCFI ignores most user-specified CFI directives
+.*40: Warning: GINSN: found unreachable code in func 'bar'
diff --git a/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s b/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s
new file mode 100644
index 00000000000..89ae29f37df
--- /dev/null
+++ b/gas/testsuite/gas/scfi/x86_64/scfi-cfg-3.s
@@ -0,0 +1,43 @@ 
+# Testcase with jmp to function instead of a call.
+# Also includes a jmp to label locally defined.
+# The CFG creation process is not expected to warn about
+# missing foo_handler_v1 or xstrdup.
+	.text
+	.globl  foo_handler
+	.type   foo_handler, @function
+foo_handler:
+	.cfi_startproc
+	movl    current_style(%rip), %eax
+	jmp     .L2
+.L2:
+	cmpl    $-1, %eax
+	je      .L5
+	testb   $4, %al
+	je      .L3
+	jmp     foo_handler_v1
+.L3:
+	xorl    %eax, %eax
+	ret
+.L5:
+	jmp     xstrdup
+	.cfi_endproc
+	.size   foo_handler, .-foo_handler
+
+# New function with unconditional jump to a label previously defined
+# in a different function.  For SCFI purposes, since .L2 in not found
+# when processing bar, a warning is issued.
+	.globl  bar
+	.type   bar, @function
+bar:
+	.cfi_startproc
+	pushq   %rbp
+	.cfi_def_cfa_offset 16
+	.cfi_offset %rbp, -16
+	movq    %rsp, %rbp
+	.cfi_def_cfa_register %rbp
+	movl    $0, %eax
+	jmp     .L2
+	popq    %rbp
+	ret
+	.cfi_endproc
+	.size   bar, .-bar
diff --git a/gas/testsuite/gas/scfi/x86_64/scfi-x86-64.exp b/gas/testsuite/gas/scfi/x86_64/scfi-x86-64.exp
index 5324af386f8..c004d99b4ac 100644
--- a/gas/testsuite/gas/scfi/x86_64/scfi-x86-64.exp
+++ b/gas/testsuite/gas/scfi/x86_64/scfi-x86-64.exp
@@ -78,6 +78,8 @@  if  { ([istarget "x86_64-*-*"] && ![istarget "x86_64-*-linux*-gnux32"]) } then {
     run_list_test "scfi-cfg-1" "--scfi=experimental --warn"
     run_dump_test "scfi-cfg-2"
     run_list_test "scfi-cfg-2" "--scfi=experimental --warn"
+    run_dump_test "scfi-cfg-3"
+    run_list_test "scfi-cfg-3" "--scfi=experimental --warn"
     run_dump_test "scfi-asm-marker-1"
     run_list_test "scfi-asm-marker-1" "--scfi=experimental --warn"
     run_dump_test "scfi-asm-marker-2"