Skip to contentSkip to conversation

x86: vsmp: avoid link error with gcc -Og

PATCH In review Arnd Bergmann 2 participants 3 messages Nov 2, 2018 – Nov 5, 2018 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86: vsmp: avoid link error with gcc -Og

When the compiler skips optimizations for ease of debugging, teh vsmp_cap_cpus() function is not eliminated from the build with CONFIG_PCI=n:

arch/x86/kernel/vsmp_64.o: In function `vsmp_cap_cpus':
vsmp_64.c:(.init.text+0x23): undefined reference to `read_pci_config'

Change the caller in a way that will always optimize this as required.

---
 arch/x86/kernel/vsmp_64.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
arch/x86/kernel/vsmp_64.c +4 -6
index 1eae5af491c2..6a2ae6964ce5 100644
@@ -146,8 +146,10 @@ static void __init detect_vsmp_box(void)
 		is_vsmp = 1;
 }
 
-static int is_vsmp_box(void)
+static int __init is_vsmp_box(void)
 {
+	detect_vsmp_box();
+
 	if (is_vsmp != -1)
 		return is_vsmp;
 	else {
@@ -157,10 +159,7 @@ static int is_vsmp_box(void)
 }
 
 #else
-static void __init detect_vsmp_box(void)
-{
-}
-static int is_vsmp_box(void)
+static __always_inline int is_vsmp_box(void)
 {
 	return 0;
 }
@@ -213,7 +212,6 @@ static void vsmp_apic_post_init(void)
 
 void __init vsmp_init(void)
 {
-	detect_vsmp_box();
 	if (!is_vsmp_box())
 		return;
 
Signature · 1 line
2.18.0
#1 lore raw

Arnd,

Quoted 3 lines · Arnd Bergmann

When the compiler skips optimizations for ease of debugging, teh vsmp_cap_cpus() function is not eliminated from the build with CONFIG_PCI=n:

thanks for that, but the code has other issues, which are addressed with this:

https://lkml.kernel.org/r/[email protected]

That should also solve the problem you've observed.

Thanks,

tglx

#2 lore raw
Quoted 8 lines · Thomas Gleixner

Arnd,

Quoted 3 lines · Arnd Bergmann

When the compiler skips optimizations for ease of debugging, teh vsmp_cap_cpus() function is not eliminated from the build with CONFIG_PCI=n:

Quoted 7 lines

thanks for that, but the code has other issues, which are addressed with this:

https://lkml.kernel.org/r/[email protected]

That should also solve the problem you've observed.

Ok, I've dropped my patch from my test tree, will let you know if I see any other issues with the new patch applied.

Arnd