Instead of leaving a trap for the next person who comes along and wants
to add something to mem_section, add an __aligned() and remove the
manual padding added for MEMCG.
Signed-off-by: Cody P Schafer <[email protected]>
---
include/linux/mmzone.h | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
---
Also, does anyone know what causes this alignment to be required here? I found
this was breaking things in a patchset I'm working on (WARNs in sysfs code
about duplicate filenames when initing mem_sections). Adding some documentation
for the reason would be appreciated.
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index 131989a..a8e8056 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -1125,9 +1125,8 @@ struct mem_section {
* section. (see memcontrol.h/page_cgroup.h about this.)
*/
struct page_cgroup *page_cgroup;
- unsigned long pad;
#endif
-};
+} __aligned(2 * sizeof(unsigned long));
#ifdef CONFIG_SPARSEMEM_EXTREME
#define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
--
1.8.3
On Thu 30 May 2013 07:14:39 AM CST, Cody P Schafer wrote:
> Instead of leaving a trap for the next person who comes along and wants
> to add something to mem_section, add an __aligned() and remove the
> manual padding added for MEMCG.
>
> Signed-off-by: Cody P Schafer <[email protected]>
> ---
> include/linux/mmzone.h | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> ---
>
> Also, does anyone know what causes this alignment to be required here? I found
> this was breaking things in a patchset I'm working on (WARNs in sysfs code
> about duplicate filenames when initing mem_sections). Adding some documentation
> for the reason would be appreciated.
Hi Cody,
I think the alignment requirement is caused by the way the
mem_section array is
organized. Basically it requires that PAGE_SIZE could be divided by
sizeof(struct mem_section).
So your change seems risky too because it should be aligned to power of
two instead
of 2 * sizeof(long).
Regards!
Gerry
>
> diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
> index 131989a..a8e8056 100644
> --- a/include/linux/mmzone.h
> +++ b/include/linux/mmzone.h
> @@ -1125,9 +1125,8 @@ struct mem_section {
> * section. (see memcontrol.h/page_cgroup.h about this.)
> */
> struct page_cgroup *page_cgroup;
> - unsigned long pad;
> #endif
> -};
> +} __aligned(2 * sizeof(unsigned long));
>
> #ifdef CONFIG_SPARSEMEM_EXTREME
> #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
On 05/29/2013 05:54 PM, Jiang Liu wrote:
> On Thu 30 May 2013 07:14:39 AM CST, Cody P Schafer wrote:
>> Also, does anyone know what causes this alignment to be required here? I found
>> this was breaking things in a patchset I'm working on (WARNs in sysfs code
>> about duplicate filenames when initing mem_sections). Adding some documentation
>> for the reason would be appreciated.
> Hi Cody,
> I think the alignment requirement is caused by the way the
> mem_section array is
> organized. Basically it requires that PAGE_SIZE could be divided by
> sizeof(struct mem_section).
> So your change seems risky too because it should be aligned to power of
> two instead
> of 2 * sizeof(long).
Well, if that's the case then this patch is wrong, and manual padding
may be the only way to go. :(
On 05/29/2013 04:14 PM, Cody P Schafer wrote:
> Instead of leaving a trap for the next person who comes along and wants
> to add something to mem_section, add an __aligned() and remove the
> manual padding added for MEMCG.
It doesn't need to be aligned technically. It needs to be a power-of-2:
http://lkml.indiana.edu/hypermail/linux/kernel/1205.2/03077.html
I'd be quite happy for someone to resurrect that patch, though. We need
a big fat comment in there.
Instead of leaving a hidden trap for the next person who comes along and
wants to add something to mem_section, add a big fat warning about it
needing to be a power-of-2, and insert a BUILD_BUG_ON() in sparse_init()
to catch mistakes.
Right now non-power-of-2 mem_sections cause a number of WARNs at boot
(which don't clearly point to the size of mem_section as an issue), but
the system limps on (temporarily, at least).
This is based upon Dave Hansen's earlier RFC where he ran into the same
issue:
"sparsemem: fix boot when SECTIONS_PER_ROOT is not power-of-2"
http://lkml.indiana.edu/hypermail/linux/kernel/1205.2/03077.html
Signed-off-by: Cody P Schafer <[email protected]>
---
Dave: Consider it resurrected.
---
include/linux/mmzone.h | 4 ++++
mm/sparse.c | 3 +++
2 files changed, 7 insertions(+)
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index 131989a..88e23f3 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -1127,6 +1127,10 @@ struct mem_section {
struct page_cgroup *page_cgroup;
unsigned long pad;
#endif
+ /*
+ * WARNING: mem_section must be a power-of-2 in size for the
+ * calculation and use of SECTION_ROOT_MASK to make sense.
+ */
};
#ifdef CONFIG_SPARSEMEM_EXTREME
diff --git a/mm/sparse.c b/mm/sparse.c
index 1c91f0d3..3194ec4 100644
--- a/mm/sparse.c
+++ b/mm/sparse.c
@@ -481,6 +481,9 @@ void __init sparse_init(void)
struct page **map_map;
#endif
+ /* see include/linux/mmzone.h 'struct mem_section' definition */
+ BUILD_BUG_ON(!is_power_of_2(sizeof(struct mem_section)));
+
/* Setup pageblock_order for HUGETLB_PAGE_SIZE_VARIABLE */
set_pageblock_order();
--
1.8.3
On 05/30/2013 11:40 AM, Cody P Schafer wrote:
> Instead of leaving a hidden trap for the next person who comes along and
> wants to add something to mem_section, add a big fat warning about it
> needing to be a power-of-2, and insert a BUILD_BUG_ON() in sparse_init()
> to catch mistakes.
>
> Right now non-power-of-2 mem_sections cause a number of WARNs at boot
> (which don't clearly point to the size of mem_section as an issue), but
> the system limps on (temporarily, at least).
>
> This is based upon Dave Hansen's earlier RFC where he ran into the same
> issue:
> "sparsemem: fix boot when SECTIONS_PER_ROOT is not power-of-2"
> http://lkml.indiana.edu/hypermail/linux/kernel/1205.2/03077.html
Thanks for doing that, Cody. At the risk of patting myself on the back:
Acked-by: Dave Hansen <[email protected]>