Re: Build regressions/improvements in v6.0-rc4

[Kees Cook]

On Mon, Sep 05, 2022 at 09:46:01AM +0200, Geert Uytterhoeven wrote:
> On Mon, 5 Sep 2022, Geert Uytterhoeven wrote:
> > JFYI, when comparing v6.0-rc4[1] to v6.0-rc3[3], the summaries are:
> >  - build errors: +3/-16
> 
>   + /kisskb/src/arch/sh/kernel/machvec.c: error: array subscript 'struct sh_machine_vector[0]' is partly outside array bounds of 'long int[1]' [-Werror=array-bounds]:  => 105:33
> 
> sh4-gcc11/sh-allyesconfig (-Werror)

Interesting -- I wonder why this suddenly appeared. I think the fix is
the common "linker addresses need to be char arrays" fix:

diff --git a/arch/sh/include/asm/sections.h b/arch/sh/include/asm/sections.h
index 8edb824049b9..0cb0ca149ac3 100644
--- a/arch/sh/include/asm/sections.h
+++ b/arch/sh/include/asm/sections.h
@@ -4,7 +4,7 @@
 
 #include <asm-generic/sections.h>
 
-extern long __machvec_start, __machvec_end;
+extern char __machvec_start[], __machvec_end[];
 extern char __uncached_start, __uncached_end;
 extern char __start_eh_frame[], __stop_eh_frame[];
 
diff --git a/arch/sh/kernel/machvec.c b/arch/sh/kernel/machvec.c
index d606679a211e..57efaf5b82ae 100644
--- a/arch/sh/kernel/machvec.c
+++ b/arch/sh/kernel/machvec.c
@@ -20,8 +20,8 @@
 #define MV_NAME_SIZE 32
 
 #define for_each_mv(mv) \
-	for ((mv) = (struct sh_machine_vector *)&__machvec_start; \
-	     (mv) && (unsigned long)(mv) < (unsigned long)&__machvec_end; \
+	for ((mv) = (struct sh_machine_vector *)__machvec_start; \
+	     (mv) && (unsigned long)(mv) < (unsigned long)__machvec_end; \
 	     (mv)++)
 
 static struct sh_machine_vector * __init get_mv_byname(const char *name)
@@ -87,8 +87,8 @@ void __init sh_mv_setup(void)
 	if (!machvec_selected) {
 		unsigned long machvec_size;
 
-		machvec_size = ((unsigned long)&__machvec_end -
-				(unsigned long)&__machvec_start);
+		machvec_size = ((unsigned long)__machvec_end -
+				(unsigned long)__machvec_start);
 
 		/*
 		 * Sanity check for machvec section alignment. Ensure
@@ -102,7 +102,7 @@ void __init sh_mv_setup(void)
 		 * vector (usually the only one) from .machvec.init.
 		 */
 		if (machvec_size >= sizeof(struct sh_machine_vector))
-			sh_mv = *(struct sh_machine_vector *)&__machvec_start;
+			sh_mv = *(struct sh_machine_vector *)__machvec_start;
 	}
 
 	pr_notice("Booting machvec: %s\n", get_system_type());

> 
>   + /kisskb/src/drivers/gpu/drm/amd/amdgpu/../display/dc/dml/dcn32/display_mode_vba_32.c: error: the frame size of 2144 bytes is larger than 2048 bytes [-Werror=frame-larger-than=]:  => 3768:1
> 
> x86_64-gcc8/x86-allmodconfig (in function dml32_ModeSupportAndSystemConfigurationFull())

This I don't know about it, but looks like a recent commit: dda4fb85e433f
Given it's a 2000 line function, I assume it could be improved! :)

>   + /kisskb/src/include/linux/fortify-string.h: error: call to '__write_overflow_field' declared with attribute warning: detected write beyond size of field (1st parameter); maybe use struct_group()? [-Werror=attribute-warning]:  => 258:25
> 
> s390x-gcc11/s390-allyesconfig (inlined from 'copy_process' at /kisskb/src/kernel/fork.c:2200:2)

This is:

        memset(&p->irqtrace, 0, sizeof(p->irqtrace));

p->irqtrace is:

        struct irqtrace_events          irqtrace;

But that's a whole object destination... why would only s390 warn?

-Kees

-- 
Kees Cook