2013-05-14 11:50:01

by majianpeng

[permalink] [raw]

Subject: [PATCH 3/3] mm/kmemleak.c: Merge the consecutive scan-areas.

---

If the scan-areas are adjacent,it can merge in order to reduce memomy.
And using pr_warn instead of pr_warning.

Signed-off-by: Jianpeng Ma <[email protected]>
---
mm/kmemleak.c | 26 +++++++++++++++++++-------
1 file changed, 19 insertions(+), 7 deletions(-)

diff --git a/mm/kmemleak.c b/mm/kmemleak.c
index f0ece93..9590a57 100644
--- a/mm/kmemleak.c
+++ b/mm/kmemleak.c
@@ -746,24 +746,36 @@ static void add_scan_area(unsigned long ptr, size_t size, gfp_t gfp)
return;
}

- area = kmem_cache_alloc(scan_area_cache, gfp_kmemleak_mask(gfp));
- if (!area) {
- pr_warning("Cannot allocate a scan area\n");
- goto out;
- }
-
spin_lock_irqsave(&object->lock, flags);
if (ptr + size > object->pointer + object->size) {
kmemleak_warn("Scan area larger than object 0x%08lx\n", ptr);
dump_object_info(object);
- kmem_cache_free(scan_area_cache, area);
goto out_unlock;
}
+ hlist_for_each_entry(area, &object->area_list, node) {
+ if (ptr + size == area->start) {
+ area->start = ptr;
+ area->size += size;
+ goto out_unlock;
+ } else if (ptr == area->start + area->size) {
+ area->size += size;
+ goto out_unlock;
+ }
+
+ }
+ spin_unlock_irqrestore(&object->lock, flags);
+
+ area = kmem_cache_alloc(scan_area_cache, gfp_kmemleak_mask(gfp));
+ if (!area) {
+ pr_warn("Cannot allocate a scan area\n");
+ goto out;
+ }

INIT_HLIST_NODE(&area->node);
area->start = ptr;
area->size = size;

+ spin_lock_irqsave(&object->lock, flags);
hlist_add_head(&area->node, &object->area_list);
out_unlock:
spin_unlock_irqrestore(&object->lock, flags);
--
1.8.3.rc1.44.gb387c77

---

Attachments:

0003-mm-kmemleak.c-Merge-the-consecutive-scan-areas.patch (1.79 kB)

2013-05-24 17:06:59

by Catalin Marinas

[permalink] [raw]

Subject: Re: [PATCH 3/3] mm/kmemleak.c: Merge the consecutive scan-areas.

---

On Tue, May 14, 2013 at 12:49:51PM +0100, majianpeng wrote:
> If the scan-areas are adjacent,it can merge in order to reduce memomy.

Have you found any significant reduction in the memory size?

What we miss though is removing an area (and I found a use-case for it).

> + hlist_for_each_entry(area, &object->area_list, node) {
> + if (ptr + size == area->start) {
> + area->start = ptr;
> + area->size += size;
> + goto out_unlock;
> + } else if (ptr == area->start + area->size) {
> + area->size += size;
> + goto out_unlock;

I prefer to keep 'goto' only for the error path. You could add a 'bool
merged' and another 'if' block for area allocation.

I'll pick the other too patches.

Thanks.

--
Catalin