diff mbox series

fix "unreplaced" warnings caused by using typeof() on inline functions

Message ID 20220624164601.99527-1-lucvoo@kernel.org (mailing list archive)
State Mainlined, archived
Headers show
Series fix "unreplaced" warnings caused by using typeof() on inline functions | expand

Commit Message

Luc Van Oostenryck June 24, 2022, 4:46 p.m. UTC 
From: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>

Currently, sparse do all its inlining at the tree level, during
constant expansion. To not mix-up the evaluation of the original
function body in case the address of an inline function is taken or
when the function can't otherwise be inlined, the statements and
symbols lists of inline functions are kept in separated fields.
Then, if the original body must be evaluated it must first be
'uninlined' to have a copy in the usual fields.

This make sense when dealing with the definition of the function.
But, when using typeof() on functions, the resulting type doesn't
refer to this definition, it's just a copy of the type and only
of the type. There shouldn't be any reasons to uninline anything.
However, the distinction between 'full function' and 'type only'
is not made during evaluation and the uninlining attempt produce
a lot of "warning: unreplaced symbol '...'" because of the lack
of a corresponding definition.

Fix this by not doing the uninlining if the symbol lack a definition.

Note: It would maybe be more appropriate for EXPR_TYPE to use
      a stripped-own version of evaluate_symbol() doing only the
      examination of the return and argument types, bypassing the
      attempt to uninline the body and evaluate the initializer and
      the statements since there is none of those for an EXPR_TYPE.

Link: https://lore.kernel.org/all/202206191726.wq70mbMK-lkp@intel.com
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
---
 evaluate.c                                    |  2 +-
 validation/inline-early/unreplaced-abstract.c | 28 +++++++++++++++++++
 validation/optim/devirtualize0.c              | 17 +++++++++++
 3 files changed, 46 insertions(+), 1 deletion(-)
 create mode 100644 validation/inline-early/unreplaced-abstract.c
 create mode 100644 validation/optim/devirtualize0.c

Comments

Andy Shevchenko July 1, 2022, 12:10 p.m. UTC | #1 
On Fri, Jun 24, 2022 at 06:46:01PM +0200, Luc Van Oostenryck wrote:
> From: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
> 
> Currently, sparse do all its inlining at the tree level, during
> constant expansion. To not mix-up the evaluation of the original
> function body in case the address of an inline function is taken or
> when the function can't otherwise be inlined, the statements and
> symbols lists of inline functions are kept in separated fields.
> Then, if the original body must be evaluated it must first be
> 'uninlined' to have a copy in the usual fields.
> 
> This make sense when dealing with the definition of the function.
> But, when using typeof() on functions, the resulting type doesn't
> refer to this definition, it's just a copy of the type and only
> of the type. There shouldn't be any reasons to uninline anything.
> However, the distinction between 'full function' and 'type only'
> is not made during evaluation and the uninlining attempt produce
> a lot of "warning: unreplaced symbol '...'" because of the lack
> of a corresponding definition.
> 
> Fix this by not doing the uninlining if the symbol lack a definition.
> 
> Note: It would maybe be more appropriate for EXPR_TYPE to use
>       a stripped-own version of evaluate_symbol() doing only the
>       examination of the return and argument types, bypassing the
>       attempt to uninline the body and evaluate the initializer and
>       the statements since there is none of those for an EXPR_TYPE.

Uwe, can we get a Debian package with this fix, it's really a tons of such
messages when compile kernel with C=1?
Uwe Kleine-König July 1, 2022, 1:31 p.m. UTC | #2 
On 7/1/22 14:10, Andy Shevchenko wrote:
> On Fri, Jun 24, 2022 at 06:46:01PM +0200, Luc Van Oostenryck wrote:
>> From: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
>>
>> Currently, sparse do all its inlining at the tree level, during
>> constant expansion. To not mix-up the evaluation of the original
>> function body in case the address of an inline function is taken or
>> when the function can't otherwise be inlined, the statements and
>> symbols lists of inline functions are kept in separated fields.
>> Then, if the original body must be evaluated it must first be
>> 'uninlined' to have a copy in the usual fields.
>>
>> This make sense when dealing with the definition of the function.
>> But, when using typeof() on functions, the resulting type doesn't
>> refer to this definition, it's just a copy of the type and only
>> of the type. There shouldn't be any reasons to uninline anything.
>> However, the distinction between 'full function' and 'type only'
>> is not made during evaluation and the uninlining attempt produce
>> a lot of "warning: unreplaced symbol '...'" because of the lack
>> of a corresponding definition.
>>
>> Fix this by not doing the uninlining if the symbol lack a definition.
>>
>> Note: It would maybe be more appropriate for EXPR_TYPE to use
>>        a stripped-own version of evaluate_symbol() doing only the
>>        examination of the return and argument types, bypassing the
>>        attempt to uninline the body and evaluate the initializer and
>>        the statements since there is none of those for an EXPR_TYPE.
> 
> Uwe, can we get a Debian package with this fix, it's really a tons of such
> messages when compile kernel with C=1?

Luc, do you consider it sensible to create a 0.6.4 variant of sparse 
that includes this fix? Are there any more patches that you'd deem 
sensible to inlucde? Does that mean it's time for a 0.6.5 release?

Best regards
Uwe
Andy Shevchenko Sept. 2, 2022, 10:02 a.m. UTC | #3 
On Fri, Jul 01, 2022 at 03:10:36PM +0300, Andy Shevchenko wrote:
> On Fri, Jun 24, 2022 at 06:46:01PM +0200, Luc Van Oostenryck wrote:
> > From: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
> > 
> > Currently, sparse do all its inlining at the tree level, during
> > constant expansion. To not mix-up the evaluation of the original
> > function body in case the address of an inline function is taken or
> > when the function can't otherwise be inlined, the statements and
> > symbols lists of inline functions are kept in separated fields.
> > Then, if the original body must be evaluated it must first be
> > 'uninlined' to have a copy in the usual fields.
> > 
> > This make sense when dealing with the definition of the function.
> > But, when using typeof() on functions, the resulting type doesn't
> > refer to this definition, it's just a copy of the type and only
> > of the type. There shouldn't be any reasons to uninline anything.
> > However, the distinction between 'full function' and 'type only'
> > is not made during evaluation and the uninlining attempt produce
> > a lot of "warning: unreplaced symbol '...'" because of the lack
> > of a corresponding definition.
> > 
> > Fix this by not doing the uninlining if the symbol lack a definition.
> > 
> > Note: It would maybe be more appropriate for EXPR_TYPE to use
> >       a stripped-own version of evaluate_symbol() doing only the
> >       examination of the return and argument types, bypassing the
> >       attempt to uninline the body and evaluate the initializer and
> >       the statements since there is none of those for an EXPR_TYPE.
> 
> Uwe, can we get a Debian package with this fix, it's really a tons of such
> messages when compile kernel with C=1?

As of today it seems Debian still has old sparse version...
diff mbox series

Patch

diff --git a/evaluate.c b/evaluate.c
index 61f59ee3908e..fe716f631987 100644
--- a/evaluate.c
+++ b/evaluate.c
@@ -3555,7 +3555,7 @@  static struct symbol *evaluate_symbol(struct symbol *sym)
 		current_fn = sym;
 
 		examine_fn_arguments(base_type);
-		if (!base_type->stmt && base_type->inline_stmt)
+		if (!base_type->stmt && base_type->inline_stmt && sym->definition)
 			uninline(sym);
 		if (base_type->stmt)
 			evaluate_statement(base_type->stmt);
diff --git a/validation/inline-early/unreplaced-abstract.c b/validation/inline-early/unreplaced-abstract.c
new file mode 100644
index 000000000000..e38cd6681f14
--- /dev/null
+++ b/validation/inline-early/unreplaced-abstract.c
@@ -0,0 +1,28 @@ 
+static inline void f0(void) { }
+static inline long f1(long a) { return a + 1;}
+
+_Static_assert([typeof(f0)] != [typeof(f1)]);
+
+
+static inline void g0(void) { }
+static inline long g1(long a) { return a + 1;}
+
+extern long goo(long a);
+long goo(long a)
+{
+	g0();
+	return g1(a);
+}
+
+_Static_assert([typeof(g0)] != [typeof(g1)]);
+
+extern long moo(long a);
+long moo(long a)
+{
+	typeof(f1) *f = g1;
+	return f(a);
+}
+
+/*
+ * check-name: unreplaced-abstract
+ */
diff --git a/validation/optim/devirtualize0.c b/validation/optim/devirtualize0.c
new file mode 100644
index 000000000000..7079e79072fe
--- /dev/null
+++ b/validation/optim/devirtualize0.c
@@ -0,0 +1,17 @@ 
+static inline long f1(long x) { return x + 1;}
+
+extern long foo(long a);
+long foo(long a)
+{
+	typeof(f1) *f = f1;
+	return f(a);
+}
+
+/*
+ * check-name: devirtualize0
+ * check-command: test-linearize -Wno-decl $file
+ * check-known-to-fail
+ *
+ * check-output-ignore
+ * check-output-excludes: call\\.
+ */