Reduce amount of time kswapd sleeps prematurely v2

Changelog since v1
o Rebase to 4.11-rc1
o Add small clarifying comment based on review

The series is unusual in that the first patch fixes one problem and introduces of other issues that are noted in the changelog. Patch 2 makes a minor modification that is worth considering on its own but leaves the kernel in a state where it behaves badly. It's not until patch 3 that there is an improvement against baseline.

This was mostly motivated by examining Chris Mason's "simoop" benchmark which puts the VM under similar pressure to HADOOP. It has been reported that the benchmark has regressed severely during the last number of releases. While I cannot reproduce all the same problems Chris experienced due to hardware limitations, there was a number of problems on a 2-socket machine with a single disk.

simoop latencies

                                         4.11.0-rc1            4.11.0-rc1
                                            vanilla          keepawake-v2
Amean    p50-Read             21670074.18 (  0.00%) 22668332.52 ( -4.61%)
Amean    p95-Read             25456267.64 (  0.00%) 26738688.00 ( -5.04%)
Amean    p99-Read             29369064.73 (  0.00%) 30991404.52 ( -5.52%)
Amean    p50-Write                1390.30 (  0.00%)      924.91 ( 33.47%)
Amean    p95-Write              412901.57 (  0.00%)     1362.62 ( 99.67%)
Amean    p99-Write             6668722.09 (  0.00%)    16854.04 ( 99.75%)
Amean    p50-Allocation          78714.31 (  0.00%)    74729.74 (  5.06%)
Amean    p95-Allocation         175533.51 (  0.00%)   101609.74 ( 42.11%)
Amean    p99-Allocation         247003.02 (  0.00%)   125765.57 ( 49.08%)

These are latencies. Read/write are threads reading fixed-size random blocks from a simulated database. The allocation latency is mmaping and faulting regions of memory. The p50, 95 and p99 reports the worst latencies for 50% of the samples, 95% and 99% respectively.

For example, the report indicates that while the test was running 99% of writes completed 99.75% faster. It's worth noting that on a UMA machine that no difference in performance with simoop was observed so milage will vary.

It's noted that there is a slight impact to read latencies but it's mostly due to IO scheduler decisions and offset by the large reduction in other latencies.

 mm/memory_hotplug.c |   2 +-
 mm/vmscan.c         | 136 ++++++++++++++++++++++++++++++----------------------
 2 files changed, 79 insertions(+), 59 deletions(-)

2.11.0

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

From: Shantanu Goel <[email protected]>

The check in prepare_kswapd_sleep needs to match the one in balance_pgdat since the latter will return as soon as any one of the zones in the classzone is above the watermark. This is specially important for higher order allocations since balance_pgdat will typically reset the order to zero relying on compaction to create the higher order pages. Without this patch, prepare_kswapd_sleep fails to wake up kcompactd since the zone balance check fails.

It was first reported against 4.9.7 that kswapd is failing to wake up kcompactd due to a mismatch in the zone balance check between balance_pgdat() and prepare_kswapd_sleep(). balance_pgdat() returns as soon as a single zone satisfies the allocation but prepare_kswapd_sleep() requires all zones to do +the same. This causes prepare_kswapd_sleep() to never succeed except in the order == 0 case and consequently, wakeup_kcompactd() is never called. For the machine that originally motivated this patch, the state of compaction from /proc/vmstat looked this way after a day and a half +of uptime:

compact_migrate_scanned 240496
compact_free_scanned 76238632
compact_isolated 123472
compact_stall 1791
compact_fail 29
compact_success 1762
compact_daemon_wake 0

After applying the patch and about 10 hours of uptime the state looks like this:

compact_migrate_scanned 59927299
compact_free_scanned 2021075136
compact_isolated 640926
compact_stall 4
compact_fail 2
compact_success 2
compact_daemon_wake 5160

Further notes from Mel that motivated him to pick this patch up and resend it;

It was observed for the simoop workload (pressures the VM similar to HADOOP) that kswapd was failing to keep ahead of direct reclaim. The investigation noted that there was a need to rationalise kswapd decisions to reclaim with kswapd decisions to sleep. With this patch on a 2-socket box, there was a 49% reduction in direct reclaim scanning.

However, the impact otherwise is extremely negative. Kswapd reclaim efficiency dropped from 98% to 76%. simoop has three latency-related metrics for read, write and allocation (an anonymous mmap and fault).

                                         4.11.0-rc1            4.11.0-rc1
                                            vanilla           fixcheck-v2
Amean    p50-Read             21670074.18 (  0.00%) 20464344.18 (  5.56%)
Amean    p95-Read             25456267.64 (  0.00%) 25721423.64 ( -1.04%)
Amean    p99-Read             29369064.73 (  0.00%) 30174230.76 ( -2.74%)
Amean    p50-Write                1390.30 (  0.00%)     1395.28 ( -0.36%)
Amean    p95-Write              412901.57 (  0.00%)    37737.74 ( 90.86%)
Amean    p99-Write             6668722.09 (  0.00%)   666489.04 ( 90.01%)
Amean    p50-Allocation          78714.31 (  0.00%)    86286.22 ( -9.62%)
Amean    p95-Allocation         175533.51 (  0.00%)   351812.27 (-100.42%)
Amean    p99-Allocation         247003.02 (  0.00%)  6291171.56 (-2447.00%)

Of greater concern is that the patch causes swapping and page writes from kswapd context rose from 0 pages to 4189753 pages during the hour the workload ran for. By and large, the patch has very bad behaviour but easily missed as the impact on a UMA machine is negligible.

This patch is included with the data in case a bisection leads to this area. This patch is also a pre-requisite for the rest of the series.

---
 mm/vmscan.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

index bc8031ef994d..4ea444142c2e 100644
@@ -3134,11 +3134,11 @@ static bool prepare_kswapd_sleep(pg_data_t *pgdat, int order, int classzone_idx)
 		if (!managed_zone(zone))
 			continue;
 
-		if (!zone_balanced(zone, order, classzone_idx))
-			return false;
+		if (zone_balanced(zone, order, classzone_idx))
+			return true;
 	}
 
-	return true;
+	return false;
 }
 
 /*

@@ -3331,7 +3331,13 @@ static void kswapd_try_to_sleep(pg_data_t *pgdat, int alloc_order, int reclaim_o
 
 	prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
 
-	/* Try to sleep for a short interval */
+	/*
+	 * Try to sleep for a short interval. Note that kcompactd will only be
+	 * woken if it is possible to sleep for a short interval. This is
+	 * deliberate on the assumption that if reclaim cannot keep an
+	 * eligible zone balanced that it's also unlikely that compaction will
+	 * succeed.
+	 */
 	if (prepare_kswapd_sleep(pgdat, reclaim_order, classzone_idx)) {
 		/*
 		 * Compaction records what page blocks it recently failed to

2.11.0

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

A pgdat tracks if recent reclaim encountered too many dirty, writeback or congested pages. The flags control whether kswapd writes pages back from reclaim context, tags pages for immediate reclaim when IO completes, whether processes block on wait_iff_congested and whether kswapd blocks when too many pages marked for immediate reclaim are encountered.

The state is cleared in a check function with side-effects. With the patch "mm, vmscan: fix zone balance check in prepare_kswapd_sleep", the timing of when the bits get cleared changed. Due to the way the check works, it'll clear the bits if ZONE_DMA is balanced for a GFP_DMA allocation because it does not account for lowmem reserves properly.

For the simoop workload, kswapd is not stalling when it should due to the premature clearing, writing pages from reclaim context like crazy and generally being unhelpful.

This patch resets the pgdat bits related to page reclaim only when kswapd is going to sleep. The comparison with simoop is then

4.11.0-rc1 4.11.0-rc1 4.11.0-rc1
vanilla fixcheck-v2 clear-v2

Amean    p50-Read             21670074.18 (  0.00%) 20464344.18 (  5.56%) 19786774.76 (  8.69%)
Amean    p95-Read             25456267.64 (  0.00%) 25721423.64 ( -1.04%) 24101956.27 (  5.32%)
Amean    p99-Read             29369064.73 (  0.00%) 30174230.76 ( -2.74%) 27691872.71 (  5.71%)
Amean    p50-Write                1390.30 (  0.00%)     1395.28 ( -0.36%)     1011.91 ( 27.22%)
Amean    p95-Write              412901.57 (  0.00%)    37737.74 ( 90.86%)    34874.98 ( 91.55%)
Amean    p99-Write             6668722.09 (  0.00%)   666489.04 ( 90.01%)   575449.60 ( 91.37%)
Amean    p50-Allocation          78714.31 (  0.00%)    86286.22 ( -9.62%)    84246.26 ( -7.03%)
Amean    p95-Allocation         175533.51 (  0.00%)   351812.27 (-100.42%)   400058.43 (-127.91%)
Amean    p99-Allocation         247003.02 (  0.00%)  6291171.56 (-2447.00%) 10905600.00 (-4315.17%)

Read latency is improved, write latency is mostly improved but allocation latency is regressed. kswapd is still reclaiming inefficiently, pages are being written back from writeback context and a host of other issues. However, given the change, it needed to be spelled out why the side-effect was moved.

---
 mm/vmscan.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)

index 4ea444142c2e..17b1afbce88e 100644
@@ -3091,17 +3091,17 @@ static bool zone_balanced(struct zone *zone, int order, int classzone_idx)
 	if (!zone_watermark_ok_safe(zone, order, mark, classzone_idx))
 		return false;
 
-	/*
-	 * If any eligible zone is balanced then the node is not considered
-	 * to be congested or dirty
-	 */
-	clear_bit(PGDAT_CONGESTED, &zone->zone_pgdat->flags);
-	clear_bit(PGDAT_DIRTY, &zone->zone_pgdat->flags);
-	clear_bit(PGDAT_WRITEBACK, &zone->zone_pgdat->flags);
-
 	return true;
 }
 
+/* Clear pgdat state for congested, dirty or under writeback. */
+static void clear_pgdat_congested(pg_data_t *pgdat)
+{
+	clear_bit(PGDAT_CONGESTED, &pgdat->flags);
+	clear_bit(PGDAT_DIRTY, &pgdat->flags);
+	clear_bit(PGDAT_WRITEBACK, &pgdat->flags);
+}
+
 /*
  * Prepare kswapd for sleeping. This verifies that there are no processes
  * waiting in throttle_direct_reclaim() and that watermarks have been met.

@@ -3134,8 +3134,10 @@ static bool prepare_kswapd_sleep(pg_data_t *pgdat, int order, int classzone_idx)
 		if (!managed_zone(zone))
 			continue;
 
-		if (zone_balanced(zone, order, classzone_idx))
+		if (zone_balanced(zone, order, classzone_idx)) {
+			clear_pgdat_congested(pgdat);
 			return true;
+		}
 	}
 
 	return false;

2.11.0

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

kswapd is woken to reclaim a node based on a failed allocation request from any eligible zone. Once reclaiming in balance_pgdat(), it will continue reclaiming until there is an eligible zone available for the zone it was woken for. kswapd tracks what zone it was recently woken for in pgdat->kswapd_classzone_idx. If it has not been woken recently, this zone will be 0.

However, the decision on whether to sleep is made on kswapd_classzone_idx which is 0 without a recent wakeup request and that classzone does not account for lowmem reserves. This allows kswapd to sleep when a low small zone such as ZONE_DMA is balanced for a GFP_DMA request even if a stream of allocations cannot use that zone. While kswapd may be woken again shortly in the near future there are two consequences -- the pgdat bits that control congestion are cleared prematurely and direct reclaim is more likely as kswapd slept prematurely.

This patch flips kswapd_classzone_idx to default to MAX_NR_ZONES (an invalid index) when there has been no recent wakeups. If there are no wakeups, it'll decide whether to sleep based on the highest possible zone available (MAX_NR_ZONES - 1). It then becomes critical that the "pgdat balanced" decisions during reclaim and when deciding to sleep are the same. If there is a mismatch, kswapd can stay awake continually trying to balance tiny zones.

simoop was used to evaluate it again. Two of the preparation patches regressed the workload so they are included as the second set of results. Otherwise this patch looks artifically excellent

4.11.0-rc1 4.11.0-rc1 4.11.0-rc1
vanilla clear-v2 keepawake-v2

Amean    p50-Read             21670074.18 (  0.00%) 19786774.76 (  8.69%) 22668332.52 ( -4.61%)
Amean    p95-Read             25456267.64 (  0.00%) 24101956.27 (  5.32%) 26738688.00 ( -5.04%)
Amean    p99-Read             29369064.73 (  0.00%) 27691872.71 (  5.71%) 30991404.52 ( -5.52%)
Amean    p50-Write                1390.30 (  0.00%)     1011.91 ( 27.22%)      924.91 ( 33.47%)
Amean    p95-Write              412901.57 (  0.00%)    34874.98 ( 91.55%)     1362.62 ( 99.67%)
Amean    p99-Write             6668722.09 (  0.00%)   575449.60 ( 91.37%)    16854.04 ( 99.75%)
Amean    p50-Allocation          78714.31 (  0.00%)    84246.26 ( -7.03%)    74729.74 (  5.06%)
Amean    p95-Allocation         175533.51 (  0.00%)   400058.43 (-127.91%)   101609.74 ( 42.11%)
Amean    p99-Allocation         247003.02 (  0.00%) 10905600.00 (-4315.17%)   125765.57 ( 49.08%)

With this patch on top, write and allocation latencies are massively improved. The read latencies are slightly impaired but it's worth noting that this is mostly due to the IO scheduler and not directly related to reclaim. The vmstats are a bit of a mix but the relevant ones are as follows;

                            4.10.0-rc7  4.10.0-rc7  4.10.0-rc7
                          mmots-20170209 clear-v1r25keepawake-v1r25
Swap Ins                             0           0           0
Swap Outs                            0         608           0
Direct pages scanned           6910672     3132699     6357298
Kswapd pages scanned          57036946    82488665    56986286
Kswapd pages reclaimed        55993488    63474329    55939113

Direct pages reclaimed 6905990 2964843 6352115
Kswapd efficiency 98% 76% 98%
Kswapd velocity 12494.375 17597.507 12488.065
Direct efficiency 99% 94% 99%
Direct velocity 1513.835 668.306 1393.148
Page writes by reclaim 0.000 4410243.000 0.000

Page writes file                     0     4409635           0
Page writes anon                     0         608           0

Page reclaim immediate 1036792 14175203 1042571

4.11.0-rc1 4.11.0-rc1 4.11.0-rc1
vanilla clear-v2 keepawake-v2

Swap Ins                             0          12           0
Swap Outs                            0         838           0
Direct pages scanned           6579706     3237270     6256811
Kswapd pages scanned          61853702    79961486    54837791
Kswapd pages reclaimed        60768764    60755788    53849586
Direct pages reclaimed         6579055     2987453     6256151
Kswapd efficiency                  98%         75%         98%
Page writes by reclaim           0.000 4389496.000       0.000
Page writes file                     0     4388658           0
Page writes anon                     0         838           0

Page reclaim immediate 1073573 14473009 982507

Swap-outs are equivalent to baseline.
Direct reclaim is reduced but not eliminated. It's worth noting
that there are two periods of direct reclaim for this workload. The first is when it switches from preparing the files for the actual test itself. It's a lot of file IO followed by a lot of allocs that reclaims heavily for a brief window. While direct reclaim is lower with clear-v2, it is due to kswapd scanning aggressively and trying to reclaim the world which is not the right thing to do. With the patches applied, there is still direct reclaim but the phase change from "creating work files" to starting multiple threads that allocate a lot of anonymous memory faster than kswapd can reclaim.
Scanning/reclaim efficiency is restored by this patch.
Page writes from reclaim context are back at 0 which is ideal. Pages immediately reclaimed after IO completes is slightly improved but it
is expected this will vary slightly.

On UMA, there is almost no change so this is not expected to be a universal win.

---
 mm/memory_hotplug.c |   2 +-
 mm/vmscan.c         | 118 +++++++++++++++++++++++++++++-----------------------
 2 files changed, 66 insertions(+), 54 deletions(-)

index 295479b792ec..edff09061e32 100644
@@ -1207,7 +1207,7 @@ static pg_data_t __ref *hotadd_new_pgdat(int nid, u64 start)
 		/* Reset the nr_zones, order and classzone_idx before reuse */
 		pgdat->nr_zones = 0;
 		pgdat->kswapd_order = 0;
-		pgdat->kswapd_classzone_idx = 0;
+		pgdat->kswapd_classzone_idx = MAX_NR_ZONES;
 	}
 
 	/* we can use NODE_DATA(nid) from here */

2.11.0

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>