Re: [PATCH 1/2] Export get_user_pages_fast

[Andrew Morton <akpm@linux-foundation.org>]

On Thu, 31 Jul 2008 18:10:57 +1000 Nick Piggin <nickpiggin@yahoo.com.au> wrote:

> On Thursday 31 July 2008 17:02, Andrew Morton wrote:
> > On Thu, 31 Jul 2008 09:43:59 +0300 Avi Kivity <avi@qumranet.com> wrote:
> 
> > > Surprised.  Why not an out-of-line function?  Would eliminate the
> > > present discussion as well.
> >
> > Excellent question!
> 
> Like this?
> 

Boy, that cleaned out a lot of crud, didn't it?

> Index: linux-2.6/include/linux/mm.h
> ===================================================================
> --- linux-2.6.orig/include/linux/mm.h	2008-07-31 17:59:04.000000000 +1000
> +++ linux-2.6/include/linux/mm.h	2008-07-31 18:01:51.000000000 +1000
> @@ -834,7 +834,6 @@ extern int mprotect_fixup(struct vm_area
>  			  struct vm_area_struct **pprev, unsigned long start,
>  			  unsigned long end, unsigned long newflags);
>  
> -#ifdef CONFIG_HAVE_GET_USER_PAGES_FAST
>  /*
>   * get_user_pages_fast provides equivalent functionality to get_user_pages,
>   * operating on current and current->mm (force=0 and doesn't return any vmas).
> @@ -848,25 +847,6 @@ extern int mprotect_fixup(struct vm_area
>  int get_user_pages_fast(unsigned long start, int nr_pages, int write,
>  			struct page **pages);
>  
> -#else
> -/*
> - * Should probably be moved to asm-generic, and architectures can include it if
> - * they don't implement their own get_user_pages_fast.
> - */
> -#define get_user_pages_fast(start, nr_pages, write, pages)	\
> -({								\
> -	struct mm_struct *mm = current->mm;			\
> -	int ret;						\
> -								\
> -	down_read(&mm->mmap_sem);				\
> -	ret = get_user_pages(current, mm, start, nr_pages,	\
> -					write, 0, pages, NULL);	\
> -	up_read(&mm->mmap_sem);					\
> -								\
> -	ret;							\
> -})
> -#endif
> -
>  /*
>   * A callback you can register to apply pressure to ageable caches.
>   *
> Index: linux-2.6/mm/memory.c
> ===================================================================
> --- linux-2.6.orig/mm/memory.c	2008-07-31 17:59:26.000000000 +1000
> +++ linux-2.6/mm/memory.c	2008-07-31 18:01:25.000000000 +1000
> @@ -1270,6 +1270,21 @@ int get_user_pages(struct task_struct *t
>  }
>  EXPORT_SYMBOL(get_user_pages);
>  
> +int __attribute__((weak)) get_user_pages_fast(unsigned long start,
> +				int nr_pages, int write, struct page **pages)
> +{
> +	struct mm_struct *mm = current->mm;
> +	int ret;
> +
> +	down_read(&mm->mmap_sem);
> +	ret = get_user_pages(current, mm, start, nr_pages,
> +					write, 0, pages, NULL);
> +	up_read(&mm->mmap_sem);
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(get_user_pages_fast);

hm, OK, I guess there's no reason why gcc should muck up the export of
a weak symbol.


>  pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
>  			spinlock_t **ptl)
>  {
> Index: linux-2.6/arch/x86/Kconfig
> ===================================================================
> --- linux-2.6.orig/arch/x86/Kconfig	2008-07-31 18:08:35.000000000 +1000
> +++ linux-2.6/arch/x86/Kconfig	2008-07-31 18:08:53.000000000 +1000
> @@ -22,7 +22,6 @@ config X86
>  	select HAVE_IDE
>  	select HAVE_OPROFILE
>  	select HAVE_IOREMAP_PROT
> -	select HAVE_GET_USER_PAGES_FAST
>  	select HAVE_KPROBES
>  	select ARCH_WANT_OPTIONAL_GPIOLIB
>  	select HAVE_KRETPROBES
> Index: linux-2.6/arch/x86/mm/Makefile
> ===================================================================
> --- linux-2.6.orig/arch/x86/mm/Makefile	2008-07-31 18:08:58.000000000 +1000
> +++ linux-2.6/arch/x86/mm/Makefile	2008-07-31 18:09:07.000000000 +1000
> @@ -1,7 +1,6 @@
>  obj-y	:=  init_$(BITS).o fault.o ioremap.o extable.o pageattr.o mmap.o \
> -	    pat.o pgtable.o
> +	    pat.o pgtable.o gup.o
>  
> -obj-$(CONFIG_HAVE_GET_USER_PAGES_FAST) += gup.o
>  obj-$(CONFIG_X86_32)		+= pgtable_32.o
>  
>  obj-$(CONFIG_HUGETLB_PAGE)	+= hugetlbpage.o

If we do this then corresponding changes should be made to the powerpc
patch upon which Ben presently sits.