On Fri, May 17, 2013 at 02:42:48PM -0700, David Daney wrote: > On 05/17/2013 02:06 PM, Sanjay Lal wrote: > >The KVM module uses the standard MIPS cache management routines, which use min_low_pfn. > >This creates and indirect dependency, requiring min_low_pfn to be exported. > > > >Signed-off-by: Sanjay Lal <sanjayl@xxxxxxxxxxx> > >--- > > arch/mips/kernel/mips_ksyms.c | 6 ++++++ > > 1 file changed, 6 insertions(+) > > > >diff --git a/arch/mips/kernel/mips_ksyms.c b/arch/mips/kernel/mips_ksyms.c > >index 6e58e97..0299472 100644 > >--- a/arch/mips/kernel/mips_ksyms.c > >+++ b/arch/mips/kernel/mips_ksyms.c > >@@ -14,6 +14,7 @@ > > #include <linux/mm.h> > > #include <asm/uaccess.h> > > #include <asm/ftrace.h> > >+#include <linux/bootmem.h> > > > > extern void *__bzero(void *__s, size_t __count); > > extern long __strncpy_from_user_nocheck_asm(char *__to, > >@@ -60,3 +61,8 @@ EXPORT_SYMBOL(invalid_pte_table); > > /* _mcount is defined in arch/mips/kernel/mcount.S */ > > EXPORT_SYMBOL(_mcount); > > #endif > >+ > >+/* The KVM module uses the standard MIPS cache functions which use > >+ * min_low_pfn, requiring it to be exported. The comment is wrong. min_low_pfn is being referenced by pfn_valid() which is implemented (simplified for purposes of this discussion) like: int pfn_valid(unsigned long pfn) { return pfn >= min_low_pfn && pfn < max_mapnr; } > >+ */ > >+EXPORT_SYMBOL(min_low_pfn); > > I think I asked this before, but I don't remember the answer: > > Why not put EXPORT_SYMBOL(min_low_pfn) in mm/bootmem.c adjacent to > where the symbol is defined? > > Cluttering up the kernel with multiple architectures all doing > architecture specific exports of the same symbol is not a clean way > of doing things. > > The second time something needs to be done, it should be factored > out into common code. pfn_valid() should return 1 if for the pfn passed as the argment a struct page exists in mem_map[] - this only affects the flatmem case. However it is possible that min_low_pfn is bigger than ARCH_PFN_OFFSET in which case pfn_valid might return 0 even though it should have returned 1. So I fixed the issue like: static inline int pfn_valid(unsigned long pfn) { return pfn >= ARCH_PFN_OFFSET && pfn < max_mapnr; } frv, ia64, metag, microblaze, score and sh are also implementing pfn_valid referencing min_low_pfn. Time to move pfn_valid() to asm-generic, fix the min_low_pfn / ARCH_PFN_OFFSET issue and while at it, harden it against multiple evaluation of its arguments? Ralf -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx"> email@xxxxxxxxx </a>