[PATCH mm v3 0/3] memcg: Slow down swap allocation as the available space gets depleted

[PATCH mm v3 0/3] memcg: Slow down swap allocation as the available space gets depleted

[Jakub Kicinski]

Tejun describes the problem as follows:

When swap runs out, there's an abrupt change in system behavior -
the anonymous memory suddenly becomes unmanageable which readily
breaks any sort of memory isolation and can bring down the whole
system. To avoid that, oomd [1] monitors free swap space and triggers
kills when it drops below the specific threshold (e.g. 15%).

While this works, it's far from ideal:
 - Depending on IO performance and total swap size, a given
   headroom might not be enough or too much.
 - oomd has to monitor swap depletion in addition to the usual
   pressure metrics and it currently doesn't consider memory.swap.max.

Solve this by adapting parts of the approach that memory.high uses -
slow down allocation as the resource gets depleted turning the
depletion behavior from abrupt cliff one to gradual degradation
observable through memory pressure metric.

[1] https://github.com/facebookincubator/oomd

v2: https://lore.kernel.org/linux-mm/20200511225516.2431921-1-kuba@kernel.org/
v1: https://lore.kernel.org/linux-mm/20200417010617.927266-1-kuba@kernel.org/

Jakub Kicinski (3):
  mm: prepare for swap over-high accounting and penalty calculation
  mm: move penalty delay clamping out of calculate_high_delay()
  mm: automatically penalize tasks with high swap use

 Documentation/admin-guide/cgroup-v2.rst |  20 +++
 include/linux/memcontrol.h              |   4 +
 mm/memcontrol.c                         | 159 ++++++++++++++++++------
 3 files changed, 143 insertions(+), 40 deletions(-)

-- 
2.25.4

[PATCH mm v3 1/3] mm: prepare for swap over-high accounting and penalty calculation

[Jakub Kicinski]

Slice the memory overage calculation logic a little bit so we can
reuse it to apply a similar penalty to the swap. The logic which
accesses the memory-specific fields (use and high values) has to
be taken out of calculate_high_delay().

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
v4: calculate max outside of calculate_overage()
---
 mm/memcontrol.c | 62 ++++++++++++++++++++++++++++---------------------
 1 file changed, 35 insertions(+), 27 deletions(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index fe4f4d96ae3e..851f4d033e8f 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -2300,41 +2300,48 @@ static void high_work_func(struct work_struct *work)
  #define MEMCG_DELAY_PRECISION_SHIFT 20
  #define MEMCG_DELAY_SCALING_SHIFT 14
 
-/*
- * Get the number of jiffies that we should penalise a mischievous cgroup which
- * is exceeding its memory.high by checking both it and its ancestors.
- */
-static unsigned long calculate_high_delay(struct mem_cgroup *memcg,
-					  unsigned int nr_pages)
+static u64 calculate_overage(unsigned long usage, unsigned long high)
 {
-	unsigned long penalty_jiffies;
-	u64 max_overage = 0;
-
-	do {
-		unsigned long usage, high;
-		u64 overage;
+	u64 overage;
 
-		usage = page_counter_read(&memcg->memory);
-		high = READ_ONCE(memcg->high);
+	if (usage <= high)
+		return 0;
 
-		if (usage <= high)
-			continue;
+	/*
+	 * Prevent division by 0 in overage calculation by acting as if
+	 * it was a threshold of 1 page
+	 */
+	high = max(high, 1UL);
 
-		/*
-		 * Prevent division by 0 in overage calculation by acting as if
-		 * it was a threshold of 1 page
-		 */
-		high = max(high, 1UL);
+	overage = usage - high;
+	overage <<= MEMCG_DELAY_PRECISION_SHIFT;
+	return div64_u64(overage, high);
+}
 
-		overage = usage - high;
-		overage <<= MEMCG_DELAY_PRECISION_SHIFT;
-		overage = div64_u64(overage, high);
+static u64 mem_find_max_overage(struct mem_cgroup *memcg)
+{
+	u64 overage, max_overage = 0;
 
-		if (overage > max_overage)
-			max_overage = overage;
+	do {
+		overage = calculate_overage(page_counter_read(&memcg->memory),
+					    READ_ONCE(memcg->high));
+		max_overage = max(overage, max_overage);
 	} while ((memcg = parent_mem_cgroup(memcg)) &&
 		 !mem_cgroup_is_root(memcg));
 
+	return max_overage;
+}
+
+/*
+ * Get the number of jiffies that we should penalise a mischievous cgroup which
+ * is exceeding its memory.high by checking both it and its ancestors.
+ */
+static unsigned long calculate_high_delay(struct mem_cgroup *memcg,
+					  unsigned int nr_pages,
+					  u64 max_overage)
+{
+	unsigned long penalty_jiffies;
+
 	if (!max_overage)
 		return 0;
 
@@ -2390,7 +2397,8 @@ void mem_cgroup_handle_over_high(void)
 	 * memory.high is breached and reclaim is unable to keep up. Throttle
 	 * allocators proactively to slow down excessive growth.
 	 */
-	penalty_jiffies = calculate_high_delay(memcg, nr_pages);
+	penalty_jiffies = calculate_high_delay(memcg, nr_pages,
+					       mem_find_max_overage(memcg));
 
 	/*
 	 * Don't sleep if the amount of jiffies this memcg owes us is so low
-- 
2.25.4

[PATCH mm v3 2/3] mm: move penalty delay clamping out of calculate_high_delay()

[Jakub Kicinski]

We will want to call calculate_high_delay() twice - once for
memory and once for swap, and we should apply the clamp value
to sum of the penalties. Clamping has to be applied outside
of calculate_high_delay().

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 mm/memcontrol.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 851f4d033e8f..b2022f98bf46 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -2365,14 +2365,7 @@ static unsigned long calculate_high_delay(struct mem_cgroup *memcg,
 	 * MEMCG_CHARGE_BATCH pages is nominal, so work out how much smaller or
 	 * larger the current charge patch is than that.
 	 */
-	penalty_jiffies = penalty_jiffies * nr_pages / MEMCG_CHARGE_BATCH;
-
-	/*
-	 * Clamp the max delay per usermode return so as to still keep the
-	 * application moving forwards and also permit diagnostics, albeit
-	 * extremely slowly.
-	 */
-	return min(penalty_jiffies, MEMCG_MAX_HIGH_DELAY_JIFFIES);
+	return penalty_jiffies * nr_pages / MEMCG_CHARGE_BATCH;
 }
 
 /*
@@ -2400,6 +2393,13 @@ void mem_cgroup_handle_over_high(void)
 	penalty_jiffies = calculate_high_delay(memcg, nr_pages,
 					       mem_find_max_overage(memcg));
 
+	/*
+	 * Clamp the max delay per usermode return so as to still keep the
+	 * application moving forwards and also permit diagnostics, albeit
+	 * extremely slowly.
+	 */
+	penalty_jiffies = min(penalty_jiffies, MEMCG_MAX_HIGH_DELAY_JIFFIES);
+
 	/*
 	 * Don't sleep if the amount of jiffies this memcg owes us is so low
 	 * that it's not even worth doing, in an attempt to be nice to those who
-- 
2.25.4

[PATCH mm v3 3/3] mm: automatically penalize tasks with high swap use

[Jakub Kicinski]

Add a memory.swap.high knob, which can be used to protect the system
from SWAP exhaustion. The mechanism used for penalizing is similar
to memory.high penalty (sleep on return to user space), but with
a less steep slope.

That is not to say that the knob itself is equivalent to memory.high.
The objective is more to protect the system from potentially buggy
tasks consuming a lot of swap and impacting other tasks, or even
bringing the whole system to stand still with complete SWAP
exhaustion. Hopefully without the need to find per-task hard
limits.

Slowing misbehaving tasks down gradually allows user space oom
killers or other protection mechanisms to react. oomd and earlyoom
already do killing based on swap exhaustion, and memory.swap.high
protection will help implement such userspace oom policies more
reliably.

Use one counter for number of pages allocated under pressure
to save struct task space and avoid two separate hierarchy
walks on the hot path.

Take the new high limit into account when determining if swap
is "full". Borrowing the explanation from Johannes:

  The idea behind "swap full" is that as long as the workload has plenty
  of swap space available and it's not changing its memory contents, it
  makes sense to generously hold on to copies of data in the swap
  device, even after the swapin. A later reclaim cycle can drop the page
  without any IO. Trading disk space for IO.

  But the only two ways to reclaim a swap slot is when they're faulted
  in and the references go away, or by scanning the virtual address space
  like swapoff does - which is very expensive (one could argue it's too
  expensive even for swapoff, it's often more practical to just reboot).

  So at some point in the fill level, we have to start freeing up swap
  slots on fault/swapin. Otherwise we could eventually run out of swap
  slots while they're filled with copies of data that is also in RAM.

  We don't want to OOM a workload because its available swap space is
  filled with redundant cache.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
--
v3:
 - count events for all groups over limit
 - add doc for high events
 - remove the magic scaling factor
 - improve commit message
v2:
 - add docs,
 - improve commit message.
---
 Documentation/admin-guide/cgroup-v2.rst | 20 ++++++
 include/linux/memcontrol.h              |  4 ++
 mm/memcontrol.c                         | 83 +++++++++++++++++++++++--
 3 files changed, 101 insertions(+), 6 deletions(-)

diff --git a/Documentation/admin-guide/cgroup-v2.rst b/Documentation/admin-guide/cgroup-v2.rst
index fed4e1d2a343..1536deb2f28e 100644
--- a/Documentation/admin-guide/cgroup-v2.rst
+++ b/Documentation/admin-guide/cgroup-v2.rst
@@ -1373,6 +1373,22 @@ PAGE_SIZE multiple when read back.
 	The total amount of swap currently being used by the cgroup
 	and its descendants.
 
+  memory.swap.high
+	A read-write single value file which exists on non-root
+	cgroups.  The default is "max".
+
+	Swap usage throttle limit.  If a cgroup's swap usage exceeds
+	this limit, all its further allocations will be throttled to
+	allow userspace to implement custom out-of-memory procedures.
+
+	This limit marks a point of no return for the cgroup. It is NOT
+	designed to manage the amount of swapping a workload does
+	during regular operation. Compare to memory.swap.max, which
+	prohibits swapping past a set amount, but lets the cgroup
+	continue unimpeded as long as other memory can be reclaimed.
+
+	Healthy workloads are not expected to reach this limit.
+
   memory.swap.max
 	A read-write single value file which exists on non-root
 	cgroups.  The default is "max".
@@ -1386,6 +1402,10 @@ PAGE_SIZE multiple when read back.
 	otherwise, a value change in this file generates a file
 	modified event.
 
+	  high
+		The number of times the cgroup's swap usage was over
+		the high threshold.
+
 	  max
 		The number of times the cgroup's swap usage was about
 		to go over the max boundary and swap allocation
diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index e0bcef180672..abf1d7aad48a 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -42,6 +42,7 @@ enum memcg_memory_event {
 	MEMCG_MAX,
 	MEMCG_OOM,
 	MEMCG_OOM_KILL,
+	MEMCG_SWAP_HIGH,
 	MEMCG_SWAP_MAX,
 	MEMCG_SWAP_FAIL,
 	MEMCG_NR_MEMORY_EVENTS,
@@ -209,6 +210,9 @@ struct mem_cgroup {
 	/* Upper bound of normal memory consumption range */
 	unsigned long high;
 
+	/* Upper bound of swap consumption range */
+	unsigned long swap_high;
+
 	/* Range enforcement for interrupt charges */
 	struct work_struct high_work;
 
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index b2022f98bf46..4fe6cebb5b4b 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -2332,6 +2332,22 @@ static u64 mem_find_max_overage(struct mem_cgroup *memcg)
 	return max_overage;
 }
 
+static u64 swap_find_max_overage(struct mem_cgroup *memcg)
+{
+	u64 overage, max_overage = 0;
+
+	do {
+		overage = calculate_overage(page_counter_read(&memcg->swap),
+					    READ_ONCE(memcg->swap_high));
+		if (overage)
+			memcg_memory_event(memcg, MEMCG_SWAP_HIGH);
+		max_overage = max(overage, max_overage);
+	} while ((memcg = parent_mem_cgroup(memcg)) &&
+		 !mem_cgroup_is_root(memcg));
+
+	return max_overage;
+}
+
 /*
  * Get the number of jiffies that we should penalise a mischievous cgroup which
  * is exceeding its memory.high by checking both it and its ancestors.
@@ -2393,6 +2409,13 @@ void mem_cgroup_handle_over_high(void)
 	penalty_jiffies = calculate_high_delay(memcg, nr_pages,
 					       mem_find_max_overage(memcg));
 
+	/*
+	 * Make the swap curve more gradual, swap can be considered "cheaper",
+	 * and is allocated in larger chunks. We want the delays to be gradual.
+	 */
+	penalty_jiffies += calculate_high_delay(memcg, nr_pages,
+						swap_find_max_overage(memcg));
+
 	/*
 	 * Clamp the max delay per usermode return so as to still keep the
 	 * application moving forwards and also permit diagnostics, albeit
@@ -2583,12 +2606,23 @@ static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
 	 * reclaim, the cost of mismatch is negligible.
 	 */
 	do {
-		if (page_counter_read(&memcg->memory) > READ_ONCE(memcg->high)) {
-			/* Don't bother a random interrupted task */
-			if (in_interrupt()) {
+		bool mem_high, swap_high;
+
+		mem_high = page_counter_read(&memcg->memory) >
+			READ_ONCE(memcg->high);
+		swap_high = page_counter_read(&memcg->swap) >
+			READ_ONCE(memcg->swap_high);
+
+		/* Don't bother a random interrupted task */
+		if (in_interrupt()) {
+			if (mem_high) {
 				schedule_work(&memcg->high_work);
 				break;
 			}
+			continue;
+		}
+
+		if (mem_high || swap_high) {
 			current->memcg_nr_pages_over_high += batch;
 			set_notify_resume(current);
 			break;
@@ -5005,6 +5039,7 @@ mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
 
 	WRITE_ONCE(memcg->high, PAGE_COUNTER_MAX);
 	memcg->soft_limit = PAGE_COUNTER_MAX;
+	WRITE_ONCE(memcg->swap_high, PAGE_COUNTER_MAX);
 	if (parent) {
 		memcg->swappiness = mem_cgroup_swappiness(parent);
 		memcg->oom_kill_disable = parent->oom_kill_disable;
@@ -5158,6 +5193,7 @@ static void mem_cgroup_css_reset(struct cgroup_subsys_state *css)
 	page_counter_set_low(&memcg->memory, 0);
 	WRITE_ONCE(memcg->high, PAGE_COUNTER_MAX);
 	memcg->soft_limit = PAGE_COUNTER_MAX;
+	WRITE_ONCE(memcg->swap_high, PAGE_COUNTER_MAX);
 	memcg_wb_domain_size_changed(memcg);
 }
 
@@ -6978,10 +7014,13 @@ bool mem_cgroup_swap_full(struct page *page)
 	if (!memcg)
 		return false;
 
-	for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg))
-		if (page_counter_read(&memcg->swap) * 2 >=
-		    READ_ONCE(memcg->swap.max))
+	for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg)) {
+		unsigned long usage = page_counter_read(&memcg->swap);
+
+		if (usage * 2 >= READ_ONCE(memcg->swap_high) ||
+		    usage * 2 >= READ_ONCE(memcg->swap.max))
 			return true;
+	}
 
 	return false;
 }
@@ -7004,6 +7043,30 @@ static u64 swap_current_read(struct cgroup_subsys_state *css,
 	return (u64)page_counter_read(&memcg->swap) * PAGE_SIZE;
 }
 
+static int swap_high_show(struct seq_file *m, void *v)
+{
+	unsigned long high = READ_ONCE(mem_cgroup_from_seq(m)->swap_high);
+
+	return seq_puts_memcg_tunable(m, high);
+}
+
+static ssize_t swap_high_write(struct kernfs_open_file *of,
+			       char *buf, size_t nbytes, loff_t off)
+{
+	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
+	unsigned long high;
+	int err;
+
+	buf = strstrip(buf);
+	err = page_counter_memparse(buf, "max", &high);
+	if (err)
+		return err;
+
+	WRITE_ONCE(memcg->swap_high, high);
+
+	return nbytes;
+}
+
 static int swap_max_show(struct seq_file *m, void *v)
 {
 	return seq_puts_memcg_tunable(m,
@@ -7031,6 +7094,8 @@ static int swap_events_show(struct seq_file *m, void *v)
 {
 	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
 
+	seq_printf(m, "high %lu\n",
+		   atomic_long_read(&memcg->memory_events[MEMCG_SWAP_HIGH]));
 	seq_printf(m, "max %lu\n",
 		   atomic_long_read(&memcg->memory_events[MEMCG_SWAP_MAX]));
 	seq_printf(m, "fail %lu\n",
@@ -7045,6 +7110,12 @@ static struct cftype swap_files[] = {
 		.flags = CFTYPE_NOT_ON_ROOT,
 		.read_u64 = swap_current_read,
 	},
+	{
+		.name = "swap.high",
+		.flags = CFTYPE_NOT_ON_ROOT,
+		.seq_show = swap_high_show,
+		.write = swap_high_write,
+	},
 	{
 		.name = "swap.max",
 		.flags = CFTYPE_NOT_ON_ROOT,
-- 
2.25.4

Re: [PATCH mm v3 3/3] mm: automatically penalize tasks with high swap use

[Shakeel Butt]

On Fri, May 15, 2020 at 1:20 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> Add a memory.swap.high knob, which can be used to protect the system
> from SWAP exhaustion. The mechanism used for penalizing is similar
> to memory.high penalty (sleep on return to user space), but with
> a less steep slope.
>
> That is not to say that the knob itself is equivalent to memory.high.
> The objective is more to protect the system from potentially buggy
> tasks consuming a lot of swap and impacting other tasks, or even
> bringing the whole system to stand still with complete SWAP
> exhaustion. Hopefully without the need to find per-task hard
> limits.
>
> Slowing misbehaving tasks down gradually allows user space oom
> killers or other protection mechanisms to react. oomd and earlyoom
> already do killing based on swap exhaustion, and memory.swap.high
> protection will help implement such userspace oom policies more
> reliably.
>
> Use one counter for number of pages allocated under pressure
> to save struct task space and avoid two separate hierarchy
> walks on the hot path.
>

The above para seems out of place. It took some time to realize you
are talking about current->memcg_nr_pages_over_high. IMO instead of
this para, a comment in code would be much better.

> Take the new high limit into account when determining if swap
> is "full". Borrowing the explanation from Johannes:
>
>   The idea behind "swap full" is that as long as the workload has plenty
>   of swap space available and it's not changing its memory contents, it
>   makes sense to generously hold on to copies of data in the swap
>   device, even after the swapin. A later reclaim cycle can drop the page
>   without any IO. Trading disk space for IO.
>
>   But the only two ways to reclaim a swap slot is when they're faulted
>   in and the references go away, or by scanning the virtual address space
>   like swapoff does - which is very expensive (one could argue it's too
>   expensive even for swapoff, it's often more practical to just reboot).
>
>   So at some point in the fill level, we have to start freeing up swap
>   slots on fault/swapin.

swap.high allows the user to force the kernel to start freeing swap
slots before half-full heuristic, right?

>   Otherwise we could eventually run out of swap
>   slots while they're filled with copies of data that is also in RAM.
>
>   We don't want to OOM a workload because its available swap space is
>   filled with redundant cache.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> --
> v3:
>  - count events for all groups over limit
>  - add doc for high events
>  - remove the magic scaling factor
>  - improve commit message
> v2:
>  - add docs,
>  - improve commit message.
> ---
>  Documentation/admin-guide/cgroup-v2.rst | 20 ++++++
>  include/linux/memcontrol.h              |  4 ++
>  mm/memcontrol.c                         | 83 +++++++++++++++++++++++--
>  3 files changed, 101 insertions(+), 6 deletions(-)
>
> diff --git a/Documentation/admin-guide/cgroup-v2.rst b/Documentation/admin-guide/cgroup-v2.rst
> index fed4e1d2a343..1536deb2f28e 100644
> --- a/Documentation/admin-guide/cgroup-v2.rst
> +++ b/Documentation/admin-guide/cgroup-v2.rst
> @@ -1373,6 +1373,22 @@ PAGE_SIZE multiple when read back.
>         The total amount of swap currently being used by the cgroup
>         and its descendants.
>
> +  memory.swap.high
> +       A read-write single value file which exists on non-root
> +       cgroups.  The default is "max".
> +
> +       Swap usage throttle limit.  If a cgroup's swap usage exceeds
> +       this limit, all its further allocations will be throttled to
> +       allow userspace to implement custom out-of-memory procedures.
> +
> +       This limit marks a point of no return for the cgroup. It is NOT
> +       designed to manage the amount of swapping a workload does
> +       during regular operation. Compare to memory.swap.max, which
> +       prohibits swapping past a set amount, but lets the cgroup
> +       continue unimpeded as long as other memory can be reclaimed.
> +
> +       Healthy workloads are not expected to reach this limit.
> +
>    memory.swap.max
>         A read-write single value file which exists on non-root
>         cgroups.  The default is "max".
> @@ -1386,6 +1402,10 @@ PAGE_SIZE multiple when read back.
>         otherwise, a value change in this file generates a file
>         modified event.
>
> +         high
> +               The number of times the cgroup's swap usage was over
> +               the high threshold.
> +
>           max
>                 The number of times the cgroup's swap usage was about
>                 to go over the max boundary and swap allocation
> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> index e0bcef180672..abf1d7aad48a 100644
> --- a/include/linux/memcontrol.h
> +++ b/include/linux/memcontrol.h
> @@ -42,6 +42,7 @@ enum memcg_memory_event {
>         MEMCG_MAX,
>         MEMCG_OOM,
>         MEMCG_OOM_KILL,
> +       MEMCG_SWAP_HIGH,
>         MEMCG_SWAP_MAX,
>         MEMCG_SWAP_FAIL,
>         MEMCG_NR_MEMORY_EVENTS,
> @@ -209,6 +210,9 @@ struct mem_cgroup {
>         /* Upper bound of normal memory consumption range */
>         unsigned long high;
>
> +       /* Upper bound of swap consumption range */
> +       unsigned long swap_high;
> +

I think it would be better to move the 'high' to the struct
page_counter i.e. memcg->memory.high and memcg->swap.high.

>         /* Range enforcement for interrupt charges */
>         struct work_struct high_work;
>
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index b2022f98bf46..4fe6cebb5b4b 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -2332,6 +2332,22 @@ static u64 mem_find_max_overage(struct mem_cgroup *memcg)
>         return max_overage;
>  }
>
> +static u64 swap_find_max_overage(struct mem_cgroup *memcg)
> +{
> +       u64 overage, max_overage = 0;
> +
> +       do {
> +               overage = calculate_overage(page_counter_read(&memcg->swap),
> +                                           READ_ONCE(memcg->swap_high));
> +               if (overage)
> +                       memcg_memory_event(memcg, MEMCG_SWAP_HIGH);
> +               max_overage = max(overage, max_overage);
> +       } while ((memcg = parent_mem_cgroup(memcg)) &&
> +                !mem_cgroup_is_root(memcg));
> +
> +       return max_overage;
> +}
> +
>  /*
>   * Get the number of jiffies that we should penalise a mischievous cgroup which
>   * is exceeding its memory.high by checking both it and its ancestors.
> @@ -2393,6 +2409,13 @@ void mem_cgroup_handle_over_high(void)
>         penalty_jiffies = calculate_high_delay(memcg, nr_pages,
>                                                mem_find_max_overage(memcg));
>
> +       /*
> +        * Make the swap curve more gradual, swap can be considered "cheaper",
> +        * and is allocated in larger chunks. We want the delays to be gradual.
> +        */
> +       penalty_jiffies += calculate_high_delay(memcg, nr_pages,
> +                                               swap_find_max_overage(memcg));
> +
>         /*
>          * Clamp the max delay per usermode return so as to still keep the
>          * application moving forwards and also permit diagnostics, albeit
> @@ -2583,12 +2606,23 @@ static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
>          * reclaim, the cost of mismatch is negligible.
>          */
>         do {
> -               if (page_counter_read(&memcg->memory) > READ_ONCE(memcg->high)) {
> -                       /* Don't bother a random interrupted task */
> -                       if (in_interrupt()) {
> +               bool mem_high, swap_high;
> +
> +               mem_high = page_counter_read(&memcg->memory) >
> +                       READ_ONCE(memcg->high);
> +               swap_high = page_counter_read(&memcg->swap) >
> +                       READ_ONCE(memcg->swap_high);
> +
> +               /* Don't bother a random interrupted task */
> +               if (in_interrupt()) {
> +                       if (mem_high) {
>                                 schedule_work(&memcg->high_work);
>                                 break;
>                         }
> +                       continue;

break?

> +               }
> +
> +               if (mem_high || swap_high) {
>                         current->memcg_nr_pages_over_high += batch;
>                         set_notify_resume(current);
>                         break;
> @@ -5005,6 +5039,7 @@ mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
>
>         WRITE_ONCE(memcg->high, PAGE_COUNTER_MAX);
>         memcg->soft_limit = PAGE_COUNTER_MAX;
> +       WRITE_ONCE(memcg->swap_high, PAGE_COUNTER_MAX);
>         if (parent) {
>                 memcg->swappiness = mem_cgroup_swappiness(parent);
>                 memcg->oom_kill_disable = parent->oom_kill_disable;
> @@ -5158,6 +5193,7 @@ static void mem_cgroup_css_reset(struct cgroup_subsys_state *css)
>         page_counter_set_low(&memcg->memory, 0);
>         WRITE_ONCE(memcg->high, PAGE_COUNTER_MAX);
>         memcg->soft_limit = PAGE_COUNTER_MAX;
> +       WRITE_ONCE(memcg->swap_high, PAGE_COUNTER_MAX);
>         memcg_wb_domain_size_changed(memcg);
>  }
>
> @@ -6978,10 +7014,13 @@ bool mem_cgroup_swap_full(struct page *page)
>         if (!memcg)
>                 return false;
>
> -       for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg))
> -               if (page_counter_read(&memcg->swap) * 2 >=
> -                   READ_ONCE(memcg->swap.max))
> +       for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg)) {
> +               unsigned long usage = page_counter_read(&memcg->swap);
> +
> +               if (usage * 2 >= READ_ONCE(memcg->swap_high) ||
> +                   usage * 2 >= READ_ONCE(memcg->swap.max))
>                         return true;
> +       }
>
>         return false;
>  }
> @@ -7004,6 +7043,30 @@ static u64 swap_current_read(struct cgroup_subsys_state *css,
>         return (u64)page_counter_read(&memcg->swap) * PAGE_SIZE;
>  }
>
> +static int swap_high_show(struct seq_file *m, void *v)
> +{
> +       unsigned long high = READ_ONCE(mem_cgroup_from_seq(m)->swap_high);
> +
> +       return seq_puts_memcg_tunable(m, high);
> +}
> +
> +static ssize_t swap_high_write(struct kernfs_open_file *of,
> +                              char *buf, size_t nbytes, loff_t off)
> +{
> +       struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
> +       unsigned long high;
> +       int err;
> +
> +       buf = strstrip(buf);
> +       err = page_counter_memparse(buf, "max", &high);
> +       if (err)
> +               return err;
> +
> +       WRITE_ONCE(memcg->swap_high, high);
> +
> +       return nbytes;
> +}
> +
>  static int swap_max_show(struct seq_file *m, void *v)
>  {
>         return seq_puts_memcg_tunable(m,
> @@ -7031,6 +7094,8 @@ static int swap_events_show(struct seq_file *m, void *v)
>  {
>         struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
>
> +       seq_printf(m, "high %lu\n",
> +                  atomic_long_read(&memcg->memory_events[MEMCG_SWAP_HIGH]));
>         seq_printf(m, "max %lu\n",
>                    atomic_long_read(&memcg->memory_events[MEMCG_SWAP_MAX]));
>         seq_printf(m, "fail %lu\n",
> @@ -7045,6 +7110,12 @@ static struct cftype swap_files[] = {
>                 .flags = CFTYPE_NOT_ON_ROOT,
>                 .read_u64 = swap_current_read,
>         },
> +       {
> +               .name = "swap.high",
> +               .flags = CFTYPE_NOT_ON_ROOT,
> +               .seq_show = swap_high_show,
> +               .write = swap_high_write,
> +       },
>         {
>                 .name = "swap.max",
>                 .flags = CFTYPE_NOT_ON_ROOT,
> --
> 2.25.4
>

Re: [PATCH mm v3 3/3] mm: automatically penalize tasks with high swap use

[Jakub Kicinski]

On Sun, 17 May 2020 06:44:52 -0700 Shakeel Butt wrote:
> > Use one counter for number of pages allocated under pressure
> > to save struct task space and avoid two separate hierarchy
> > walks on the hot path.
> 
> The above para seems out of place. It took some time to realize you
> are talking about current->memcg_nr_pages_over_high. IMO instead of
> this para, a comment in code would be much better.

Where would you like to see the comment? In struct task or where
counter is bumped?

> > Take the new high limit into account when determining if swap
> > is "full". Borrowing the explanation from Johannes:
> >
> >   The idea behind "swap full" is that as long as the workload has plenty
> >   of swap space available and it's not changing its memory contents, it
> >   makes sense to generously hold on to copies of data in the swap
> >   device, even after the swapin. A later reclaim cycle can drop the page
> >   without any IO. Trading disk space for IO.
> >
> >   But the only two ways to reclaim a swap slot is when they're faulted
> >   in and the references go away, or by scanning the virtual address space
> >   like swapoff does - which is very expensive (one could argue it's too
> >   expensive even for swapoff, it's often more practical to just reboot).
> >
> >   So at some point in the fill level, we have to start freeing up swap
> >   slots on fault/swapin.  
> 
> swap.high allows the user to force the kernel to start freeing swap
> slots before half-full heuristic, right?

I'd say that the definition of full is extended to include swap.high.

Re: [PATCH mm v3 3/3] mm: automatically penalize tasks with high swap use

[Shakeel Butt]

On Mon, May 18, 2020 at 12:42 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Sun, 17 May 2020 06:44:52 -0700 Shakeel Butt wrote:
> > > Use one counter for number of pages allocated under pressure
> > > to save struct task space and avoid two separate hierarchy
> > > walks on the hot path.
> >
> > The above para seems out of place. It took some time to realize you
> > are talking about current->memcg_nr_pages_over_high. IMO instead of
> > this para, a comment in code would be much better.
>
> Where would you like to see the comment? In struct task or where
> counter is bumped?
>

I think the place where the counter is bumped.

> > > Take the new high limit into account when determining if swap
> > > is "full". Borrowing the explanation from Johannes:
> > >
> > >   The idea behind "swap full" is that as long as the workload has plenty
> > >   of swap space available and it's not changing its memory contents, it
> > >   makes sense to generously hold on to copies of data in the swap
> > >   device, even after the swapin. A later reclaim cycle can drop the page
> > >   without any IO. Trading disk space for IO.
> > >
> > >   But the only two ways to reclaim a swap slot is when they're faulted
> > >   in and the references go away, or by scanning the virtual address space
> > >   like swapoff does - which is very expensive (one could argue it's too
> > >   expensive even for swapoff, it's often more practical to just reboot).
> > >
> > >   So at some point in the fill level, we have to start freeing up swap
> > >   slots on fault/swapin.
> >
> > swap.high allows the user to force the kernel to start freeing swap
> > slots before half-full heuristic, right?
>
> I'd say that the definition of full is extended to include swap.high.

Re: [PATCH mm v3 3/3] mm: automatically penalize tasks with high swap use

[Jakub Kicinski]

On Sun, 17 May 2020 06:44:52 -0700 Shakeel Butt wrote:
> > @@ -2583,12 +2606,23 @@ static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
> >          * reclaim, the cost of mismatch is negligible.
> >          */
> >         do {
> > -               if (page_counter_read(&memcg->memory) > READ_ONCE(memcg->high)) {
> > -                       /* Don't bother a random interrupted task */
> > -                       if (in_interrupt()) {
> > +               bool mem_high, swap_high;
> > +
> > +               mem_high = page_counter_read(&memcg->memory) >
> > +                       READ_ONCE(memcg->high);
> > +               swap_high = page_counter_read(&memcg->swap) >
> > +                       READ_ONCE(memcg->swap_high);
> > +
> > +               /* Don't bother a random interrupted task */
> > +               if (in_interrupt()) {
> > +                       if (mem_high) {
> >                                 schedule_work(&memcg->high_work);
> >                                 break;
> >                         }
> > +                       continue;  
> 
> break?

On a closer look I think continue is correct. In irq we only care 
about mem_high, because there's nothing we can do in a work context 
to penalize swap. So the loop is shortened.

> > +               }
> > +
> > +               if (mem_high || swap_high) {
> >                         current->memcg_nr_pages_over_high += batch;
> >                         set_notify_resume(current);
> >                         break;

Re: [PATCH mm v3 3/3] mm: automatically penalize tasks with high swap use

[Shakeel Butt]

On Mon, May 18, 2020 at 5:42 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Sun, 17 May 2020 06:44:52 -0700 Shakeel Butt wrote:
> > > @@ -2583,12 +2606,23 @@ static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
> > >          * reclaim, the cost of mismatch is negligible.
> > >          */
> > >         do {
> > > -               if (page_counter_read(&memcg->memory) > READ_ONCE(memcg->high)) {
> > > -                       /* Don't bother a random interrupted task */
> > > -                       if (in_interrupt()) {
> > > +               bool mem_high, swap_high;
> > > +
> > > +               mem_high = page_counter_read(&memcg->memory) >
> > > +                       READ_ONCE(memcg->high);
> > > +               swap_high = page_counter_read(&memcg->swap) >
> > > +                       READ_ONCE(memcg->swap_high);
> > > +
> > > +               /* Don't bother a random interrupted task */
> > > +               if (in_interrupt()) {
> > > +                       if (mem_high) {
> > >                                 schedule_work(&memcg->high_work);
> > >                                 break;
> > >                         }
> > > +                       continue;
> >
> > break?
>
> On a closer look I think continue is correct. In irq we only care
> about mem_high, because there's nothing we can do in a work context
> to penalize swap. So the loop is shortened.
>

Yes, you are right.