[v2] Only add '-fwrapv' to gcc CFLAGS for gcc 4.2 and later

Commit Message

Frans Pop July 10, 2009, 8:05 p.m. UTC

On Friday 10 July 2009, Frans Pop wrote:
> The discussion on issue looks to have died, but it has bitten Debian
> stable ("Lenny") [1] as it causes init to die on s390 after a kernel
> update.
>
> Here's a possible patch. The exact gcc version to check for is still a
> bit open I guess. For the s390 issue I've confirmed that 4.2.4 is OK,
> but for safety and because of Andrew's comment above I've set the test
> for 4.3 in the patch.

Here's an updated patch as I found the gcc version check was incorrect 
(0430 should have been 0403; sorry).

I've now changed the check to allow -fwrapv for gcc 4.2 as that has been 
shown to work and because of the consideration mentioned in my previous 
mail.

---
From: Frans Pop <[e-mail]>
Subject: Only add '-fwrapv' to gcc CFLAGS for gcc 4.2 and later

This flag has been shown to cause init to segfault for kernels
compiled with gcc-4.1. gcc version 4.2.4 has been shown to be OK.

This fixes http://bugzilla.kernel.org/show_bug.cgi?id=13012.

Reported-by: Barry K. Nathan <[e-mail]>
Signed-off-by: Frans Pop <[e-mail]>

 KBUILD_CFLAGS	+= $(call cc-option,-fno-dwarf2-cfi-asm)
--
To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

Frans Pop July 17, 2009, 10:43 p.m. UTC | #1

On Saturday 18 July 2009, Sam Ravnborg wrote:
> >  # disable invalid "can't wrap" optimizations for signed / pointers
> > -KBUILD_CFLAGS	+= $(call cc-option,-fwrapv)
> > +KBUILD_CFLAGS  += $(shell if [ $(call cc-version) -ge 0402 ]; then \
> > +		    echo $(call cc-option,-fwrapv); fi ;)
>
> This would be simpler if you use:

That's now academic as Linus decided on a different fix.

> # cc-ifversion
> # Usage:  EXTRA_CFLAGS += $(call cc-ifversion, -lt, 0402, -O1)
> cc-ifversion = $(shell [ $(call cc-version, $(CC)) $(1) $(2) ] && echo $(3))
>
> We have only one user at the moment so I understand why you missed it.

:-)

I based my patch on arch/x86/Makefile:
35:        KBUILD_CFLAGS += $(shell if [ $(call cc-version) -lt 0400 ] ; then \
36:                echo $(call cc-option,-fno-unit-at-a-time); fi ;)

Guess that could be improved to use cc-ifversion then.

And a quick git grep gives a few other potential candidates:
arch/ia64/Makefile:44:ifeq ($(call cc-version),0304)
arch/parisc/Makefile:129:       @if test "$(call cc-version)" -lt "0303"; then \
arch/powerpc/Makefile:80:GCC_BROKEN_VEC := $(shell if [ $(call cc-version) -lt 0400 ] ; then echo "y"; fi)
arch/powerpc/Makefile:219:      @if test "$(call cc-version)" = "0304" ; then \
arch/um/Makefile-i386:38:KBUILD_CFLAGS += $(shell if [ $(call cc-version) -lt 0400 ] ; then \

Cheers,
FJP
--
To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Frans Pop July 23, 2009, 12:46 p.m. UTC | #2

On Saturday 18 July 2009, Sam Ravnborg wrote:
> > I based my patch on arch/x86/Makefile:
> > 35:        KBUILD_CFLAGS += $(shell if [ $(call cc-version) -lt 0400 ] ; then \
> > 36:                echo $(call cc-option,-fno-unit-at-a-time); fi ;)
> >
> > Guess that could be improved to use cc-ifversion then.
>
> Yes, please...

I've got patches for x86, ia64, powerpc and um, but have a couple of
questions before I submit them.

Can the patches go through your kbuild tree or should they go through the
arch trees (they will be CCed of course)?

In my patches I've left the cc-option check where it existed, so for x86
you get (tested to work):
-        KBUILD_CFLAGS += $(shell if [ $(call cc-version) -lt 0400 ] ; then \
-                echo $(call cc-option,-fno-unit-at-a-time); fi ;)
+        KBUILD_CFLAGS += $(call cc-ifversion, -lt, 0400, \
+                               $(call cc-option,-fno-unit-at-a-time))

Or is it safe to simplify that to just:
+        KBUILD_CFLAGS += $(call cc-ifversion, -lt, 0400, -fno-unit-at-a-time)
?

> > And a quick git grep gives a few other potential candidates:
> > arch/parisc/Makefile:129:       @if test "$(call cc-version)" -lt "0303"; then \

I left this one unchanged as it does not add a compiler option; instead it
displays an "unsupported' message.
--
To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/Makefile b/Makefile
index 0aeec59..2519fde 100644
--- a/Makefile
+++ b/Makefile
@@ -565,7 +565,8 @@  KBUILD_CFLAGS += $(call 
cc-option,-Wdeclaration-after-statement,)
 KBUILD_CFLAGS += $(call cc-option,-Wno-pointer-sign,)
 
 # disable invalid "can't wrap" optimizations for signed / pointers
-KBUILD_CFLAGS	+= $(call cc-option,-fwrapv)
+KBUILD_CFLAGS  += $(shell if [ $(call cc-version) -ge 0402 ]; then \
+		    echo $(call cc-option,-fwrapv); fi ;)
 
 # revert to pre-gcc-4.4 behaviour of .eh_frame