[PATCH] [RFC] vmscan.c: add a sysctl entry for controlling memory reclaim IO congestion_wait length

From: Lin Feng
Date: Tue Sep 17 2019 - 08:19:12 EST

Next message: Andrey Konovalov: "Re: KASAN: slab-out-of-bounds Read in ld_usb_read (2)"
Previous message: kbuild test robot: "Re: [PATCH 3/3] dmaengine: sf-pdma: add platform DMA support for HiFive Unleashed A00"
Next in thread: Matthew Wilcox: "Re: [PATCH] [RFC] vmscan.c: add a sysctl entry for controlling memory reclaim IO congestion_wait length"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

This sysctl is named as mm_reclaim_congestion_wait_jiffies, default to
HZ/10 as unchanged to old codes.

It is in jiffies unit and can be set in range between [1, 100], so
refers to CONFIG_HZ before tuning.

In direct and background(kswapd) pages reclaim paths both may fall into
calling msleep(100) or congestion_wait(HZ/10) or wait_iff_congested(HZ/10)
while under IO pressure, and the sleep length is hard-coded and the later
two will introduce 100ms iowait length per time.

So if pages reclaim is relatively active in some circumstances such as high
order pages reappings, it's possible to see a lot of iowait introduced by
congestion_wait(HZ/10) and wait_iff_congested(HZ/10).

The 100ms sleep length is proper if the backing drivers are slow like
traditionnal rotation disks. While if the backing drivers are high-end
storages such as high iops ssds or even faster drivers, the high iowait
inroduced by pages reclaim is really misleading, because the storage IO
utils seen by iostat is quite low, in this case the congestion_wait time
modified to 1ms is likely enough for high-end ssds.

Another benifit is that it's potentially shorter the direct reclaim blocked
time when kernel falls into sync reclaim path, which may improve user
applications response time.

All ssds box is a trend, so introduce this sysctl entry for making a way
to relieving the concerns of system administrators.

Tested:
1. Before this patch:

top - 10:10:40 up 8 days, 16:22, 4 users, load average: 2.21, 2.15, 2.10
Tasks: 718 total, 5 running, 712 sleeping, 0 stopped, 1 zombie
Cpu0 : 0.3%us, 3.4%sy, 0.0%ni, 95.3%id, 1.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu1 : 1.4%us, 1.7%sy, 0.0%ni, 95.2%id, 0.0%wa, 0.0%hi, 1.7%si, 0.0%st
Cpu2 : 4.7%us, 3.3%sy, 0.0%ni, 91.0%id, 0.0%wa, 0.0%hi, 1.0%si, 0.0%st
Cpu3 : 7.0%us, 3.7%sy, 0.0%ni, 87.7%id, 1.0%wa, 0.0%hi, 0.7%si, 0.0%st
Cpu4 : 1.0%us, 2.0%sy, 0.0%ni, 96.3%id, 0.0%wa, 0.0%hi, 0.7%si, 0.0%st
Cpu5 : 1.0%us, 2.0%sy, 0.0%ni, 1.7%id, 95.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu6 : 1.0%us, 1.3%sy, 0.0%ni, 97.3%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu7 : 1.3%us, 1.0%sy, 0.0%ni, 97.7%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu8 : 4.3%us, 1.3%sy, 0.0%ni, 94.3%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu9 : 0.7%us, 0.7%sy, 0.0%ni, 98.3%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu10 : 0.7%us, 1.0%sy, 0.0%ni, 98.3%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu11 : 1.0%us, 1.0%sy, 0.0%ni, 97.7%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu12 : 3.0%us, 1.0%sy, 0.0%ni, 95.3%id, 0.3%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu13 : 0.3%us, 1.3%sy, 0.0%ni, 88.6%id, 9.4%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu14 : 3.3%us, 2.3%sy, 0.0%ni, 93.7%id, 0.3%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu15 : 6.4%us, 3.0%sy, 0.0%ni, 90.2%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu16 : 2.7%us, 1.7%sy, 0.0%ni, 95.3%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu17 : 1.0%us, 1.7%sy, 0.0%ni, 97.3%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu18 : 1.3%us, 1.0%sy, 0.0%ni, 97.0%id, 0.3%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu19 : 4.3%us, 1.7%sy, 0.0%ni, 86.0%id, 7.7%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu20 : 0.7%us, 1.3%sy, 0.0%ni, 97.7%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu21 : 0.3%us, 1.7%sy, 0.0%ni, 50.2%id, 47.5%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu22 : 0.7%us, 0.7%sy, 0.0%ni, 98.7%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu23 : 0.7%us, 0.7%sy, 0.0%ni, 98.3%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st

2. After this patch and set mm_reclaim_congestion_wait_jiffies to 1:

top - 10:12:19 up 8 days, 16:24, 4 users, load average: 1.32, 1.93, 2.03
Tasks: 724 total, 2 running, 721 sleeping, 0 stopped, 1 zombie
Cpu0 : 4.4%us, 3.0%sy, 0.0%ni, 90.3%id, 1.3%wa, 0.0%hi, 1.0%si, 0.0%st
Cpu1 : 2.1%us, 1.4%sy, 0.0%ni, 93.5%id, 0.7%wa, 0.0%hi, 2.4%si, 0.0%st
Cpu2 : 2.7%us, 1.0%sy, 0.0%ni, 96.3%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu3 : 1.0%us, 1.0%sy, 0.0%ni, 97.7%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu4 : 0.7%us, 1.0%sy, 0.0%ni, 97.7%id, 0.3%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu5 : 1.0%us, 0.7%sy, 0.0%ni, 97.7%id, 0.3%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu6 : 1.7%us, 1.0%sy, 0.0%ni, 97.3%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu7 : 2.0%us, 0.7%sy, 0.0%ni, 94.3%id, 2.7%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu8 : 2.0%us, 0.7%sy, 0.0%ni, 97.0%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu9 : 0.7%us, 1.0%sy, 0.0%ni, 97.7%id, 0.7%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu10 : 0.3%us, 0.3%sy, 0.0%ni, 99.3%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu11 : 0.7%us, 0.3%sy, 0.0%ni, 99.0%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu12 : 0.7%us, 1.0%sy, 0.0%ni, 98.0%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu13 : 0.0%us, 0.3%sy, 0.0%ni, 99.3%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu14 : 1.7%us, 0.7%sy, 0.0%ni, 97.3%id, 0.3%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu15 : 4.3%us, 1.0%sy, 0.0%ni, 94.3%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu16 : 1.7%us, 1.3%sy, 0.0%ni, 96.3%id, 0.0%wa, 0.0%hi, 0.7%si, 0.0%st
Cpu17 : 2.0%us, 1.3%sy, 0.0%ni, 96.3%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu18 : 0.3%us, 0.3%sy, 0.0%ni, 99.3%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu19 : 1.0%us, 1.0%sy, 0.0%ni, 97.6%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu20 : 1.3%us, 0.7%sy, 0.0%ni, 97.0%id, 0.7%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu21 : 0.7%us, 0.7%sy, 0.0%ni, 98.3%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
Cpu22 : 1.0%us, 1.0%sy, 0.0%ni, 98.0%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Cpu23 : 0.7%us, 0.3%sy, 0.0%ni, 98.3%id, 0.0%wa, 0.0%hi, 0.7%si, 0.0%st

Signed-off-by: Lin Feng <linf@xxxxxxxxxx>
---
Documentation/admin-guide/sysctl/vm.rst | 17 +++++++++++++++++
kernel/sysctl.c | 10 ++++++++++
mm/vmscan.c | 12 +++++++++---
3 files changed, 36 insertions(+), 3 deletions(-)

diff --git a/Documentation/admin-guide/sysctl/vm.rst b/Documentation/admin-guide/sysctl/vm.rst
index 64aeee1009ca..e4dd83731ecf 100644
--- a/Documentation/admin-guide/sysctl/vm.rst
+++ b/Documentation/admin-guide/sysctl/vm.rst
@@ -837,6 +837,23 @@ than the high water mark in a zone.
The default value is 60.

+mm_reclaim_congestion_wait_jiffies
+==========
+
+This control is used to define how long kernel will wait/sleep while
+system memory is under pressure and memroy reclaim is relatively active.
+Lower values will decrease the kernel wait/sleep time.
+
+It's suggested to lower this value on high-end box that system is under memory
+pressure but with low storage IO utils and high CPU iowait, which could also
+potentially decrease user application response time in this case.
+
+Keep this control as it were if your box are not above case.
+
+The default value is HZ/10, which is of equal value to 100ms independ of how
+many HZ is defined.
+
+
unprivileged_userfaultfd
========================

diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 078950d9605b..064a3da04789 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -114,6 +114,7 @@ extern int pid_max;
extern int pid_max_min, pid_max_max;
extern int percpu_pagelist_fraction;
extern int latencytop_enabled;
+extern int mm_reclaim_congestion_wait_jiffies;
extern unsigned int sysctl_nr_open_min, sysctl_nr_open_max;
#ifndef CONFIG_MMU
extern int sysctl_nr_trim_pages;
@@ -1413,6 +1414,15 @@ static struct ctl_table vm_table[] = {
.extra1 = SYSCTL_ZERO,
.extra2 = &one_hundred,
},
+ {
+ .procname = "mm_reclaim_congestion_wait_jiffies",
+ .data = &mm_reclaim_congestion_wait_jiffies,
+ .maxlen = sizeof(mm_reclaim_congestion_wait_jiffies),
+ .mode = 0644,
+ .proc_handler = proc_dointvec_minmax,
+ .extra1 = &SYSCTL_ONE,
+ .extra2 = &one_hundred,
+ },
#ifdef CONFIG_HUGETLB_PAGE
{
.procname = "nr_hugepages",
diff --git a/mm/vmscan.c b/mm/vmscan.c
index a6c5d0b28321..8c19afdcff95 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -165,6 +165,12 @@ struct scan_control {
* From 0 .. 100. Higher means more swappy.
*/
int vm_swappiness = 60;
+
+/*
+ * From 0 .. 100. Lower means shorter memory reclaim IO congestion wait time.
+ */
+int mm_reclaim_congestion_wait_jiffies = HZ / 10;
+
/*
* The total number of pages which are beyond the high watermark within all
* zones.
@@ -1966,7 +1972,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
return 0;

/* wait a bit for the reclaimer. */
- msleep(100);
+ msleep(jiffies_to_msecs(mm_reclaim_congestion_wait_jiffies));
stalled = true;

/* We are about to die and free our memory. Return now. */
@@ -2788,7 +2794,7 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
* faster than they are written so also forcibly stall.
*/
if (sc->nr.immediate)
- congestion_wait(BLK_RW_ASYNC, HZ/10);
+ congestion_wait(BLK_RW_ASYNC, mm_reclaim_congestion_wait_jiffies);
}

/*
@@ -2807,7 +2813,7 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
*/
if (!sc->hibernation_mode && !current_is_kswapd() &&
current_may_throttle() && pgdat_memcg_congested(pgdat, root))
- wait_iff_congested(BLK_RW_ASYNC, HZ/10);
+ wait_iff_congested(BLK_RW_ASYNC, mm_reclaim_congestion_wait_jiffies);

} while (should_continue_reclaim(pgdat, sc->nr_reclaimed - nr_reclaimed,
sc->nr_scanned - nr_scanned, sc));
--
2.20.1

Next message: Andrey Konovalov: "Re: KASAN: slab-out-of-bounds Read in ld_usb_read (2)"
Previous message: kbuild test robot: "Re: [PATCH 3/3] dmaengine: sf-pdma: add platform DMA support for HiFive Unleashed A00"
Next in thread: Matthew Wilcox: "Re: [PATCH] [RFC] vmscan.c: add a sysctl entry for controlling memory reclaim IO congestion_wait length"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]