Currently PAGE_OWNER depends on CONFIG_X86. This appears to be due to
pfn_to_page() being called in an inappropriate for many memory models
and the presense of memory holes. This patch ensures that pfn_valid()
and pfn_valid_within() is called at the appropriate places and the offsets
correctly updated so that PAGE_OWNER is safe on any architecture.
In situations where CONFIG_HOLES_IN_ZONES is set (IA64 with VIRTUAL_MEM_MAP),
there may be cases where pages allocated within a MAX_ORDER_NR_PAGES block
of pages may not be displayed in /proc/page_owner if the hole is at the
start of the block. Addressing this would be quite complex, perform slowly
and is of no clear benefit.
Once PAGE_OWNER is allowed on all architectures, the statistics for grouping
pages by mobility that declare how many pageblocks contain mixed page types
becomes optionally available on all arches.
This patch was tested successfully on x86, x86_64, ppc64 and IA64 machines.
Signed-off-by: Mel Gorman <[email protected]>
Acked-by: Andy Whitcroft <[email protected]>
---
fs/proc/proc_misc.c | 31 ++++++++++++++++++++++++-------
lib/Kconfig.debug | 2 +-
2 files changed, 25 insertions(+), 8 deletions(-)
diff -rup -X /usr/src/patchset-0.6/bin//dontdiff linux-2.6.22-rc4-mm2-clean/fs/proc/proc_misc.c linux-2.6.22-rc4-mm2-005_pageowner_anyarch/fs/proc/proc_misc.c
--- linux-2.6.22-rc4-mm2-clean/fs/proc/proc_misc.c 2007-06-07 14:11:20.000000000 +0100
+++ linux-2.6.22-rc4-mm2-005_pageowner_anyarch/fs/proc/proc_misc.c 2007-06-08 13:34:36.000000000 +0100
@@ -756,18 +756,35 @@ read_page_owner(struct file *file, char
struct page *page;
char *kbuf, *modname;
const char *symname;
- int ret = 0, next_idx = 1;
+ int ret = 0;
char namebuf[128];
unsigned long offset = 0, symsize;
int i;
ssize_t num_written = 0;
int blocktype = 0, pagetype = 0;
+ page = NULL;
pfn = min_low_pfn + *ppos;
- page = pfn_to_page(pfn);
+
+ /* Find a valid PFN or the start of a MAX_ORDER_NR_PAGES area */
+ while (!pfn_valid(pfn) && (pfn & (MAX_ORDER_NR_PAGES - 1)) != 0)
+ pfn++;
+
+ /* Find an allocated page */
for (; pfn < max_pfn; pfn++) {
- if (!pfn_valid(pfn))
+ /*
+ * If the new page is in a new MAX_ORDER_NR_PAGES area,
+ * validate the area as existing, skip it if not
+ */
+ if ((pfn & (MAX_ORDER_NR_PAGES - 1)) == 0 && !pfn_valid(pfn)) {
+ pfn += MAX_ORDER_NR_PAGES - 1;
continue;
+ }
+
+ /* Check for holes within a MAX_ORDER area */
+ if (!pfn_valid_within(pfn))
+ continue;
+
page = pfn_to_page(pfn);
/* Catch situations where free pages have a bad ->order */
@@ -776,16 +793,16 @@ read_page_owner(struct file *file, char
"PageOwner info inaccurate for PFN %lu\n",
pfn);
- if (page->order >= 0)
+ /* Stop search if page is allocated and has trace info */
+ if (page->order >= 0 && page->trace[0])
break;
-
- next_idx++;
}
if (!pfn_valid(pfn))
return 0;
- *ppos += next_idx;
+ /* Record the next PFN to read in the file offset */
+ *ppos = (pfn - min_low_pfn) + 1;
kbuf = kmalloc(count, GFP_KERNEL);
if (!kbuf)
diff -rup -X /usr/src/patchset-0.6/bin//dontdiff linux-2.6.22-rc4-mm2-clean/lib/Kconfig.debug linux-2.6.22-rc4-mm2-005_pageowner_anyarch/lib/Kconfig.debug
--- linux-2.6.22-rc4-mm2-clean/lib/Kconfig.debug 2007-06-07 14:11:21.000000000 +0100
+++ linux-2.6.22-rc4-mm2-005_pageowner_anyarch/lib/Kconfig.debug 2007-06-08 13:34:36.000000000 +0100
@@ -49,7 +49,7 @@ config UNUSED_SYMBOLS
config PAGE_OWNER
bool "Track page owner"
- depends on DEBUG_KERNEL && X86
+ depends on DEBUG_KERNEL
help
This keeps track of what call chain is the owner of a page, may
help to find bare alloc_page(s) leaks. Eats a fair amount of memory.
--
Mel Gorman
Part-time Phd Student Linux Technology Center
University of Limerick IBM Dublin Software Lab
On Fri, 8 Jun 2007, Mel Gorman wrote:
> In situations where CONFIG_HOLES_IN_ZONES is set (IA64 with VIRTUAL_MEM_MAP),
> there may be cases where pages allocated within a MAX_ORDER_NR_PAGES block
> of pages may not be displayed in /proc/page_owner if the hole is at the
> start of the block. Addressing this would be quite complex, perform slowly
> and is of no clear benefit.
Note that CONFIG_HOLES_IN_ZONES and IA64 VIRTUAL_MEM_MAP may be going
away. Andy Whitcroft has a patchset that implements virtual memmap
support under sparse and that would allow us to get rid of this.
On Fri, 8 Jun 2007 13:53:49 +0100
[email protected] (Mel Gorman) wrote:
> Currently PAGE_OWNER depends on CONFIG_X86. This appears to be due to
> pfn_to_page() being called in an inappropriate for many memory models
> and the presense of memory holes. This patch ensures that pfn_valid()
> and pfn_valid_within() is called at the appropriate places and the offsets
> correctly updated so that PAGE_OWNER is safe on any architecture.
>
> In situations where CONFIG_HOLES_IN_ZONES is set (IA64 with VIRTUAL_MEM_MAP),
> there may be cases where pages allocated within a MAX_ORDER_NR_PAGES block
> of pages may not be displayed in /proc/page_owner if the hole is at the
> start of the block. Addressing this would be quite complex, perform slowly
> and is of no clear benefit.
>
> Once PAGE_OWNER is allowed on all architectures, the statistics for grouping
> pages by mobility that declare how many pageblocks contain mixed page types
> becomes optionally available on all arches.
>
> This patch was tested successfully on x86, x86_64, ppc64 and IA64 machines.
I'm kinda mystified about how you successfully tested this on ppc64 and
ia64. They don't assemble and execute i386 opcodes?
--- a/mm/page_alloc.c~allow-page_owner-to-be-set-on-any-architecture-fix
+++ a/mm/page_alloc.c
@@ -1498,13 +1498,15 @@ static inline void __stack_trace(struct
#endif
}
-static inline void set_page_owner(struct page *page,
- unsigned int order, unsigned int gfp_mask)
+static void set_page_owner(struct page *page, unsigned int order,
+ unsigned int gfp_mask)
{
- unsigned long address, bp;
+ unsigned long address;
+ unsigned long bp = 0;
#ifdef CONFIG_X86_64
asm ("movq %%rbp, %0" : "=r" (bp) : );
-#else
+#endif
+#ifdef CONFIG_X86_32
asm ("movl %%ebp, %0" : "=r" (bp) : );
#endif
page->order = (int) order;
_
that'll make it build, but it won't work...