[PATCH] x86/mm/pat: Do a small optimization when dump PAT memtype list

[Pan Xinhui]

From: Pan Xinhui <xinhuix....@intel.com>

There are many nodes in the PAT memtype rb-tree. When we dump this tree
we call kzalloc every time to copy nodes. Actually these kzalloc are not
necessary. Lets do a optimization now.

Create an *entry* in memtype_seq_start(), and free it in
memtype_seq_stop(). These two callback functions are alwasys called in
pair. Also because memtype_seq_show() is usually used only for
outputing. Seems memtype_seq_show is not the best place to free the
*entry*

And reuse the *entry* in memtype_seq_next() to enhance the performance.

Signed-off-by: Pan Xinhui <xinhuix....@intel.com>
---
 arch/x86/mm/pat.c | 28 ++++++++++++----------------
 1 file changed, 12 insertions(+), 16 deletions(-)

diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c
index 268b2c8..6302119 100644
--- a/arch/x86/mm/pat.c
+++ b/arch/x86/mm/pat.c
@@ -1001,45 +1001,42 @@ EXPORT_SYMBOL_GPL(pgprot_writethrough);
 
 #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_X86_PAT)
 
-static struct memtype *memtype_get_idx(loff_t pos)
+static struct memtype *memtype_get_idx(struct memtype *entry, loff_t pos)
 {
-       struct memtype *print_entry;
        int ret;
 
-       print_entry  = kzalloc(sizeof(struct memtype), GFP_KERNEL);
-       if (!print_entry)
-               return NULL;
-
        spin_lock(&memtype_lock);
-       ret = rbt_memtype_copy_nth_element(print_entry, pos);
+       ret = rbt_memtype_copy_nth_element(entry, pos);
        spin_unlock(&memtype_lock);
 
-       if (!ret) {
-               return print_entry;
-       } else {
-               kfree(print_entry);
-               return NULL;
-       }
+       return ret ? NULL : entry;
 }
 
 static void *memtype_seq_start(struct seq_file *seq, loff_t *pos)
 {
+       struct memtype *entry;
+
        if (*pos == 0) {
                ++*pos;
                seq_puts(seq, "PAT memtype list:\n");
        }
 
-       return memtype_get_idx(*pos);
+       entry = kzalloc(sizeof(*entry), GFP_KERNEL);
+       seq->private = entry;
+       if (!entry)
+               return NULL;
+       return memtype_get_idx(entry, *pos);
 }
 
 static void *memtype_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 {
        ++*pos;
-       return memtype_get_idx(*pos);
+       return memtype_get_idx((struct memtype *)v, *pos);
 }
 
 static void memtype_seq_stop(struct seq_file *seq, void *v)
 {
+       kfree(seq->private);
 }
 
 static int memtype_seq_show(struct seq_file *seq, void *v)
@@ -1048,7 +1045,6 @@ static int memtype_seq_show(struct seq_file *seq, void *v)
 
        seq_printf(seq, "%s @ 0x%Lx-0x%Lx\n", cattr_name(print_entry->type),
                        print_entry->start, print_entry->end);
-       kfree(print_entry);
 
        return 0;
 }
-- 
1.9.1
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/