The patch titled
Subject: mm, thp: khugepaged should scan when sleep value is written
has been removed from the -mm tree. Its filename was
mm-thp-khugepaged-should-scan-when-sleep-value-is-written.patch
This patch was dropped because it was merged into mainline or a subsystem tree
------------------------------------------------------
From: David Rientjes <rientjes@xxxxxxxxxx>
Subject: mm, thp: khugepaged should scan when sleep value is written
If a large value is written to scan_sleep_millisecs, for example, that
period must lapse before khugepaged will wake up for periodic collapsing.
If this value is tuned to 1 day, for example, and then re-tuned to its
default 10s, khugepaged will still wait for a day before scanning again.
This patch causes khugepaged to wakeup immediately when the value is
changed and then sleep until that value is rewritten or the new value
lapses.
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1605181453200.4786@xxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: David Rientjes <rientjes@xxxxxxxxxx>
Cc: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>
Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/huge_memory.c | 19 ++++++++++++++++---
1 file changed, 16 insertions(+), 3 deletions(-)
diff -puN mm/huge_memory.c~mm-thp-khugepaged-should-scan-when-sleep-value-is-written mm/huge_memory.c
--- a/mm/huge_memory.c~mm-thp-khugepaged-should-scan-when-sleep-value-is-written
+++ a/mm/huge_memory.c
@@ -89,6 +89,7 @@ static unsigned int khugepaged_full_scan
static unsigned int khugepaged_scan_sleep_millisecs __read_mostly = 10000;
/* during fragmentation poll the hugepage allocator once every minute */
static unsigned int khugepaged_alloc_sleep_millisecs __read_mostly = 60000;
+static unsigned long khugepaged_sleep_expire;
static struct task_struct *khugepaged_thread __read_mostly;
static DEFINE_MUTEX(khugepaged_mutex);
static DEFINE_SPINLOCK(khugepaged_mm_lock);
@@ -467,6 +468,7 @@ static ssize_t scan_sleep_millisecs_stor
return -EINVAL;
khugepaged_scan_sleep_millisecs = msecs;
+ khugepaged_sleep_expire = 0;
wake_up_interruptible(&khugepaged_wait);
return count;
@@ -494,6 +496,7 @@ static ssize_t alloc_sleep_millisecs_sto
return -EINVAL;
khugepaged_alloc_sleep_millisecs = msecs;
+ khugepaged_sleep_expire = 0;
wake_up_interruptible(&khugepaged_wait);
return count;
@@ -2791,15 +2794,25 @@ static void khugepaged_do_scan(void)
put_page(hpage);
}
+static bool khugepaged_should_wakeup(void)
+{
+ return kthread_should_stop() ||
+ time_after_eq(jiffies, khugepaged_sleep_expire);
+}
+
static void khugepaged_wait_work(void)
{
if (khugepaged_has_work()) {
- if (!khugepaged_scan_sleep_millisecs)
+ const unsigned long scan_sleep_jiffies =
+ msecs_to_jiffies(khugepaged_scan_sleep_millisecs);
+
+ if (!scan_sleep_jiffies)
return;
+ khugepaged_sleep_expire = jiffies + scan_sleep_jiffies;
wait_event_freezable_timeout(khugepaged_wait,
- kthread_should_stop(),
- msecs_to_jiffies(khugepaged_scan_sleep_millisecs));
+ khugepaged_should_wakeup(),
+ scan_sleep_jiffies);
return;
}
_
Patches currently in -mm which might be from rientjes@xxxxxxxxxx are
mm-migrate-increment-fail-count-on-enomem.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
