lkml.org 
[lkml]   [2009]   [Jul]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
Subject[PATCH 1/1] pagemap clear_refs: modify to specify anon or mapped vma clearing
This patch makes the clear_refs proc interface a bit more versatile.
It adds support for clearing anonymous pages, file mapped pages or both.

The clear_refs entry is used to reset the Referenced bits on virtual and
physical pages associated with a process.
echo 1 > /proc/PID/clear_refs clears all pages associated with the process
echo 2 > /proc/PID/clear_refs clears anonymous pages only
echo 3 > /proc/PID/clear_refs clears file mapped pages only
Any other value written to the proc entry will clear all pages.

Selective clearing the pages has a measurable impact on performance as it
limits the number of page walks. We have been using this interface and this
adds flexibility to the user user space application implementing the reference
clearing.

Signed-off-by: Jared Hulbert (jaredeh@gmail.com)
Signed-off-by: Moussa A. Ba (moussa.a.ba@gmail.com)
-------
Documentation/filesystems/proc.txt | 7 +++++++
fs/proc/task_mmu.c | 29 +++++++++++++++++++++++++----
2 files changed, 32 insertions(+), 4 deletions(-)

--- a/fs/proc/task_mmu.c 2009-07-21 14:30:01.000000000 -0700
+++ b/fs/proc/task_mmu.c 2009-07-27 11:46:05.000000000 -0700
@@ -462,6 +462,27 @@
return 0;
}

+static void walk_vma_area(struct mm_walk *this_walk,
+ struct vm_area_struct *vma, int type)
+{
+
+ /* Writing 2 to /proc/pid/clear_refs will clear all Anonymous
+ * pages.
+ *
+ * Writing 3 to /proc/pid/clear_refs will clear all file mapped
+ * pages.
+ *
+ * Writing any other value including 1 will clear all pages
+ */
+ if (is_vm_hugetlb_page(vma))
+ return;
+ if (type == 2 && vma->vm_file)
+ return;
+ if (type == 3 && !vma->vm_file)
+ return;
+ walk_page_range(vma->vm_start, vma->vm_end, this_walk);
+}
+
static ssize_t clear_refs_write(struct file *file, const char __user * buf,
size_t count, loff_t * ppos)
{
@@ -469,13 +490,15 @@
char buffer[PROC_NUMBUF], *end;
struct mm_struct *mm;
struct vm_area_struct *vma;
+ int type;

memset(buffer, 0, sizeof(buffer));
if (count > sizeof(buffer) - 1)
count = sizeof(buffer) - 1;
if (copy_from_user(buffer, buf, count))
return -EFAULT;
- if (!simple_strtol(buffer, &end, 0))
+ type = strict_strtol(buffer, &end, 0);
+ if (!type)
return -EINVAL;
if (*end == '\n')
end++;
@@ -491,9 +514,7 @@
down_read(&mm->mmap_sem);
for (vma = mm->mmap; vma; vma = vma->vm_next) {
clear_refs_walk.private = vma;
- if (!is_vm_hugetlb_page(vma))
- walk_page_range(vma->vm_start, vma->vm_end,
- &clear_refs_walk);
+ walk_vma_area(&clear_refs_walk, vma, type);
}
flush_tlb_mm(mm);
up_read(&mm->mmap_sem);
--- a/Documentation/filesystems/proc.txt 2009-07-20 17:29:11.000000000 -0700
+++ b/Documentation/filesystems/proc.txt 2009-07-27 12:08:34.000000000 -0700
@@ -375,6 +375,13 @@
This file is only present if the CONFIG_MMU kernel configuration option is
enabled.

+The clear_refs entry is used to reset the Referenced bits on virtual and physical
+pages associated with a process.
+echo 1 > /proc/PID/clear_refs clears all pages associated with the process
+echo 2 > /proc/PID/clear_refs clears anonymous pages only
+echo 3 > /proc/PID/clear_refs clears file mapped pages only
+Any other value written to the proc entry will clear all pages.
+
1.2 Kernel data
---------------


\
 
 \ /
  Last update: 2009-07-27 22:21    [W:0.109 / U:0.552 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site