Skip to contentSkip to conversation

Making more drivers compile on x86

RFC New Matthew Wilcox 1 participant 1 message Sep 10, 2014 x86

lore
Share & tools
lore
#0 lore raw

RFC 1/1 Making more drivers compile on x86

In Chicago, there was a brief discussion about making more drivers compile on architectures where there was no possible way they could run (since the relevant hardware does not exist for those platforms).

I have recently needed to compile-test axonram and dcssblk, which rely on hardware unique to powerpc and s390, respectively. As promised in the discussion, here is the patch I use to do that. Since it was for my own purposes, I didn't do anything fancy like wrap the faked parts in #ifdef COMPILE_TEST.

Hopefully this will be useful grist for the mill. Looking the patch over, if I were proposing something to be merged, I think I'd make much more use of prototypes, and less of macros. It would probably also make sense to put extmem in asm-generic, rather than replicate it for each architecture that wants to have a more "full" compile test.

arch/x86/include/asm/extmem.h new file +10 -0
@@ -0,0 +1,10 @@
+#define SEGMENT_EXCLUSIVE 1
+#define SEG_TYPE_ER 2
+#define SEGMENT_SHARED 3
+#define SEG_TYPE_SR 4
+#define SEG_TYPE_SC 5
+#define segment_warning(x, y) do { } while (0)
+int segment_load(char *x, int y, unsigned long *z, unsigned long *a);
+#define segment_unload(x) do { } while (0)
+#define segment_modify_shared(x, y) 0
+#define segment_save(x) do { } while (0)
arch/x86/include/asm/prom.h +5 -0
index fbeb06e..dc6191f 100644
@@ -37,6 +37,11 @@ static inline void x86_dtb_init(void) { }
 #define of_ioapic 0
 #endif
 
+#define _PAGE_NO_CACHE 1
+#define NO_IRQ 2
+extern int of_address_to_resource(void *x, int y, struct resource *z);
+#define irq_of_parse_and_map(x, y) 0
+
 extern char cmd_line[COMMAND_LINE_SIZE];
 
 #endif /* __ASSEMBLY__ */