Re: [PATCH] mm/damon: Add option to monitor only writes

[SeongJae Park <sj@kernel.org>]

Hi Pedro,


Thank you for this patch!

On Thu, 3 Feb 2022 10:12:36 -0300 Pedro Demarchi Gomes <pedrodemargomes@gmail.com> wrote:

> When "writes" is written to /sys/kernel/debug/damon/counter_type damon will
> monitor only writes.

I think this would be better to be implemented as another monitoring primitive
based on the virtual address space monitoring primitives, e.g., vaddr-writes?
Then the implementation would be simpler and changes to other files will be
minimized.  For the user space interface, we could use a prefix to target_ids
debugfs file.  For example,

    # echo "vaddr-writes $(pidof workload)" > $debugfs/damon/target_ids


> This patch also adds the actions mergeable and unmergeable to damos schemes.
> These actions are used by KSM as explained in [1].

Please do that in a separate patch, and also update the document
(Documentation/admin-guide/mm/damon/usage.rst).  And, what's the expected usage
of the action and benefits?


Thanks,
SJ

> 
> [1] https://www.kernel.org/doc/html/latest/admin-guide/mm/ksm.html#controlling-ksm-with-madvise
> 
> Signed-off-by: Pedro Demarchi Gomes <pedrodemargomes@gmail.com>
> ---
>  include/linux/damon.h        | 13 ++++++
>  include/trace/events/damon.h | 10 ++--
>  mm/damon/core.c              | 91 +++++++++++++++++++++++++-----------
>  mm/damon/dbgfs.c             | 72 +++++++++++++++++++++++++++-
>  mm/damon/prmtv-common.c      | 88 ++++++++++++++++++++++++++++++++++
>  mm/damon/prmtv-common.h      |  5 ++
>  mm/damon/vaddr.c             | 80 ++++++++++++++++++++++++++++---
>  7 files changed, 318 insertions(+), 41 deletions(-)
> 
> diff --git a/include/linux/damon.h b/include/linux/damon.h
> index b4d4be3cc987..9efe89bbcec8 100644
> --- a/include/linux/damon.h
> +++ b/include/linux/damon.h
> @@ -44,11 +44,13 @@ struct damon_region {
>  	struct damon_addr_range ar;
>  	unsigned long sampling_addr;
>  	unsigned int nr_accesses;
> +	unsigned int nr_writes;
>  	struct list_head list;
>  
>  	unsigned int age;
>  /* private: Internal value for age calculation. */
>  	unsigned int last_nr_accesses;
> +	unsigned int last_nr_writes;
>  };
>  
>  /**
> @@ -88,6 +90,8 @@ enum damos_action {
>  	DAMOS_PAGEOUT,
>  	DAMOS_HUGEPAGE,
>  	DAMOS_NOHUGEPAGE,
> +	DAMOS_MERGEABLE,
> +	DAMOS_UNMERGEABLE,
>  	DAMOS_STAT,		/* Do nothing but only record the stat */
>  };
>  
> @@ -185,6 +189,11 @@ struct damos_watermarks {
>  	bool activated;
>  };
>  
> +enum damos_counter_type {
> +	DAMOS_NUMBER_ACCESSES,
> +	DAMOS_NUMBER_WRITES,
> +};
> +
>  /**
>   * struct damos - Represents a Data Access Monitoring-based Operation Scheme.
>   * @min_sz_region:	Minimum size of target regions.
> @@ -223,6 +232,9 @@ struct damos {
>  	unsigned long max_sz_region;
>  	unsigned int min_nr_accesses;
>  	unsigned int max_nr_accesses;
> +	unsigned int min_nr_writes;
> +	unsigned int max_nr_writes;
> +	enum damos_counter_type counter_type;
>  	unsigned int min_age_region;
>  	unsigned int max_age_region;
>  	enum damos_action action;
> @@ -386,6 +398,7 @@ struct damon_ctx {
>  	struct damon_primitive primitive;
>  	struct damon_callback callback;
>  
> +	enum damos_counter_type counter_type;
>  	unsigned long min_nr_regions;
>  	unsigned long max_nr_regions;
>  	struct list_head adaptive_targets;
> diff --git a/include/trace/events/damon.h b/include/trace/events/damon.h
> index 2f422f4f1fb9..45573f421707 100644
> --- a/include/trace/events/damon.h
> +++ b/include/trace/events/damon.h
> @@ -11,17 +11,17 @@
>  
>  TRACE_EVENT(damon_aggregated,
>  
> -	TP_PROTO(struct damon_target *t, struct damon_region *r,
> +	TP_PROTO(struct damon_ctx *c, struct damon_target *t, struct damon_region *r,
>  		unsigned int nr_regions),
>  
> -	TP_ARGS(t, r, nr_regions),
> +	TP_ARGS(c, t, r, nr_regions),
>  
>  	TP_STRUCT__entry(
>  		__field(unsigned long, target_id)
>  		__field(unsigned int, nr_regions)
>  		__field(unsigned long, start)
>  		__field(unsigned long, end)
> -		__field(unsigned int, nr_accesses)
> +		__field(unsigned int, nr)
>  	),
>  
>  	TP_fast_assign(
> @@ -29,12 +29,12 @@ TRACE_EVENT(damon_aggregated,
>  		__entry->nr_regions = nr_regions;
>  		__entry->start = r->ar.start;
>  		__entry->end = r->ar.end;
> -		__entry->nr_accesses = r->nr_accesses;
> +		__entry->nr = c->counter_type == DAMOS_NUMBER_WRITES ? r->nr_writes : r->nr_accesses;
>  	),
>  
>  	TP_printk("target_id=%lu nr_regions=%u %lu-%lu: %u",
>  			__entry->target_id, __entry->nr_regions,
> -			__entry->start, __entry->end, __entry->nr_accesses)
> +			__entry->start, __entry->end, __entry->nr)
>  );
>  
>  #endif /* _TRACE_DAMON_H */
> diff --git a/mm/damon/core.c b/mm/damon/core.c
> index e92497895202..e827231366db 100644
> --- a/mm/damon/core.c
> +++ b/mm/damon/core.c
> @@ -45,10 +45,12 @@ struct damon_region *damon_new_region(unsigned long start, unsigned long end)
>  	region->ar.start = start;
>  	region->ar.end = end;
>  	region->nr_accesses = 0;
> +	region->nr_writes = 0;
>  	INIT_LIST_HEAD(&region->list);
>  
>  	region->age = 0;
>  	region->last_nr_accesses = 0;
> +	region->last_nr_writes = 0;
>  
>  	return region;
>  }
> @@ -89,7 +91,7 @@ void damon_destroy_region(struct damon_region *r, struct damon_target *t)
>  
>  struct damos *damon_new_scheme(
>  		unsigned long min_sz_region, unsigned long max_sz_region,
> -		unsigned int min_nr_accesses, unsigned int max_nr_accesses,
> +		unsigned int min_nr, unsigned int max_nr,
>  		unsigned int min_age_region, unsigned int max_age_region,
>  		enum damos_action action, struct damos_quota *quota,
>  		struct damos_watermarks *wmarks)
> @@ -101,8 +103,10 @@ struct damos *damon_new_scheme(
>  		return NULL;
>  	scheme->min_sz_region = min_sz_region;
>  	scheme->max_sz_region = max_sz_region;
> -	scheme->min_nr_accesses = min_nr_accesses;
> -	scheme->max_nr_accesses = max_nr_accesses;
> +	scheme->max_nr_writes = max_nr;
> +	scheme->min_nr_writes = min_nr;
> +	scheme->min_nr_accesses = min_nr;
> +	scheme->max_nr_accesses = max_nr;
>  	scheme->min_age_region = min_age_region;
>  	scheme->max_age_region = max_age_region;
>  	scheme->action = action;
> @@ -221,6 +225,7 @@ struct damon_ctx *damon_new_ctx(void)
>  	ctx->sample_interval = 5 * 1000;
>  	ctx->aggr_interval = 100 * 1000;
>  	ctx->primitive_update_interval = 60 * 1000 * 1000;
> +	ctx->counter_type = DAMOS_NUMBER_ACCESSES;
>  
>  	ktime_get_coarse_ts64(&ctx->last_aggregation);
>  	ctx->last_primitive_update = ctx->last_aggregation;
> @@ -535,9 +540,11 @@ static void kdamond_reset_aggregated(struct damon_ctx *c)
>  		struct damon_region *r;
>  
>  		damon_for_each_region(r, t) {
> -			trace_damon_aggregated(t, r, damon_nr_regions(t));
> +			trace_damon_aggregated(c, t, r, damon_nr_regions(t));
>  			r->last_nr_accesses = r->nr_accesses;
>  			r->nr_accesses = 0;
> +			r->last_nr_writes = r->nr_writes;
> +			r->nr_writes = 0;
>  		}
>  	}
>  }
> @@ -546,21 +553,27 @@ static void damon_split_region_at(struct damon_ctx *ctx,
>  		struct damon_target *t, struct damon_region *r,
>  		unsigned long sz_r);
>  
> -static bool __damos_valid_target(struct damon_region *r, struct damos *s)
> +static bool __damos_valid_target(struct damon_ctx *ctx, struct damon_region *r, struct damos *s)
>  {
>  	unsigned long sz;
> -
>  	sz = r->ar.end - r->ar.start;
> -	return s->min_sz_region <= sz && sz <= s->max_sz_region &&
> -		s->min_nr_accesses <= r->nr_accesses &&
> -		r->nr_accesses <= s->max_nr_accesses &&
> -		s->min_age_region <= r->age && r->age <= s->max_age_region;
> +
> +	if (ctx->counter_type == DAMOS_NUMBER_WRITES)
> +		return s->min_sz_region <= sz && sz <= s->max_sz_region &&
> +			s->min_nr_writes <= r->nr_writes &&
> +			r->nr_writes <= s->max_nr_writes &&
> +			s->min_age_region <= r->age && r->age <= s->max_age_region;
> +	else
> +		return s->min_sz_region <= sz && sz <= s->max_sz_region &&
> +			s->min_nr_accesses <= r->nr_accesses &&
> +			r->nr_accesses <= s->max_nr_accesses &&
> +			s->min_age_region <= r->age && r->age <= s->max_age_region;
>  }
>  
>  static bool damos_valid_target(struct damon_ctx *c, struct damon_target *t,
>  		struct damon_region *r, struct damos *s)
>  {
> -	bool ret = __damos_valid_target(r, s);
> +	bool ret = __damos_valid_target(c, r, s);
>  
>  	if (!ret || !s->quota.esz || !c->primitive.get_scheme_score)
>  		return ret;
> @@ -707,7 +720,7 @@ static void kdamond_apply_schemes(struct damon_ctx *c)
>  		memset(quota->histogram, 0, sizeof(quota->histogram));
>  		damon_for_each_target(t, c) {
>  			damon_for_each_region(r, t) {
> -				if (!__damos_valid_target(r, s))
> +				if (!__damos_valid_target(c, r, s))
>  					continue;
>  				score = c->primitive.get_scheme_score(
>  						c, t, r, s);
> @@ -738,13 +751,18 @@ static void kdamond_apply_schemes(struct damon_ctx *c)
>  /*
>   * Merge two adjacent regions into one region
>   */
> -static void damon_merge_two_regions(struct damon_target *t,
> +static void damon_merge_two_regions(struct damon_ctx *c, struct damon_target *t,
>  		struct damon_region *l, struct damon_region *r)
>  {
>  	unsigned long sz_l = sz_damon_region(l), sz_r = sz_damon_region(r);
>  
> -	l->nr_accesses = (l->nr_accesses * sz_l + r->nr_accesses * sz_r) /
> +	if (c->counter_type == DAMOS_NUMBER_WRITES)
> +		l->nr_writes = (l->nr_writes * sz_l + r->nr_writes * sz_r) /
>  			(sz_l + sz_r);
> +	else
> +		l->nr_accesses = (l->nr_accesses * sz_l + r->nr_accesses * sz_r) /
> +				(sz_l + sz_r);
> +
>  	l->age = (l->age * sz_l + r->age * sz_r) / (sz_l + sz_r);
>  	l->ar.end = r->ar.end;
>  	damon_destroy_region(r, t);
> @@ -759,23 +777,39 @@ static void damon_merge_two_regions(struct damon_target *t,
>   * thres	'->nr_accesses' diff threshold for the merge
>   * sz_limit	size upper limit of each region
>   */
> -static void damon_merge_regions_of(struct damon_target *t, unsigned int thres,
> +static void damon_merge_regions_of(struct damon_ctx *c, struct damon_target *t, unsigned int thres,
>  				   unsigned long sz_limit)
>  {
>  	struct damon_region *r, *prev = NULL, *next;
>  
> -	damon_for_each_region_safe(r, next, t) {
> -		if (diff_of(r->nr_accesses, r->last_nr_accesses) > thres)
> -			r->age = 0;
> -		else
> -			r->age++;
> -
> -		if (prev && prev->ar.end == r->ar.start &&
> -		    diff_of(prev->nr_accesses, r->nr_accesses) <= thres &&
> -		    sz_damon_region(prev) + sz_damon_region(r) <= sz_limit)
> -			damon_merge_two_regions(t, prev, r);
> -		else
> -			prev = r;
> +	if (c->counter_type == DAMOS_NUMBER_WRITES) {
> +		damon_for_each_region_safe(r, next, t) {
> +			if (diff_of(r->nr_writes, r->last_nr_writes) > thres)
> +				r->age = 0;
> +			else
> +				r->age++;
> +
> +			if (prev && prev->ar.end == r->ar.start &&
> +				diff_of(prev->nr_writes, r->nr_writes) <= thres &&
> +				sz_damon_region(prev) + sz_damon_region(r) <= sz_limit)
> +				damon_merge_two_regions(c, t, prev, r);
> +			else
> +				prev = r;
> +		}
> +	} else {
> +		damon_for_each_region_safe(r, next, t) {
> +			if (diff_of(r->nr_accesses, r->last_nr_accesses) > thres)
> +				r->age = 0;
> +			else
> +				r->age++;
> +
> +			if (prev && prev->ar.end == r->ar.start &&
> +				diff_of(prev->nr_accesses, r->nr_accesses) <= thres &&
> +				sz_damon_region(prev) + sz_damon_region(r) <= sz_limit)
> +				damon_merge_two_regions(c, t, prev, r);
> +			else
> +				prev = r;
> +		}
>  	}
>  }
>  
> @@ -796,7 +830,7 @@ static void kdamond_merge_regions(struct damon_ctx *c, unsigned int threshold,
>  	struct damon_target *t;
>  
>  	damon_for_each_target(t, c)
> -		damon_merge_regions_of(t, threshold, sz_limit);
> +		damon_merge_regions_of(c, t, threshold, sz_limit);
>  }
>  
>  /*
> @@ -819,6 +853,7 @@ static void damon_split_region_at(struct damon_ctx *ctx,
>  
>  	new->age = r->age;
>  	new->last_nr_accesses = r->last_nr_accesses;
> +	new->last_nr_writes = r->last_nr_writes;
>  
>  	damon_insert_region(new, r, damon_next_region(r), t);
>  }
> diff --git a/mm/damon/dbgfs.c b/mm/damon/dbgfs.c
> index ad65436756af..6cf2501148f5 100644
> --- a/mm/damon/dbgfs.c
> +++ b/mm/damon/dbgfs.c
> @@ -166,6 +166,8 @@ static bool damos_action_valid(int action)
>  	case DAMOS_PAGEOUT:
>  	case DAMOS_HUGEPAGE:
>  	case DAMOS_NOHUGEPAGE:
> +	case DAMOS_MERGEABLE:
> +	case DAMOS_UNMERGEABLE:
>  	case DAMOS_STAT:
>  		return true;
>  	default:
> @@ -477,6 +479,66 @@ static ssize_t dbgfs_init_regions_read(struct file *file, char __user *buf,
>  	return len;
>  }
>  
> +static ssize_t dbgfs_counter_type_write(struct file *file,
> +		const char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct damon_ctx *ctx = file->private_data;
> +	char *kbuf;
> +	ssize_t ret;
> +
> +	mutex_lock(&ctx->kdamond_lock);
> +	if (ctx->kdamond) {
> +		mutex_unlock(&ctx->kdamond_lock);
> +		ret = -EBUSY;
> +		goto out;
> +	}
> +	ret = count;
> +	kbuf = user_input_str(buf, count, ppos);
> +	if (IS_ERR(kbuf))
> +		return PTR_ERR(kbuf);
> +
> +	if (!strncmp(kbuf, "writes\n", count))
> +		ctx->counter_type = DAMOS_NUMBER_WRITES;
> +	else
> +		ctx->counter_type = DAMOS_NUMBER_ACCESSES;
> +
> +	mutex_unlock(&ctx->kdamond_lock);
> +out:
> +	kfree(kbuf);
> +	return ret;
> +}
> +
> +static ssize_t dbgfs_counter_type_read(struct file *file, char __user *buf,
> +		size_t count, loff_t *ppos)
> +{
> +	struct damon_ctx *ctx = file->private_data;
> +	char *kbuf;
> +	ssize_t len;
> +
> +	kbuf = kmalloc(count, GFP_KERNEL | __GFP_NOWARN);
> +	if (!kbuf)
> +		return -ENOMEM;
> +
> +	mutex_lock(&ctx->kdamond_lock);
> +	if (ctx->kdamond) {
> +		mutex_unlock(&ctx->kdamond_lock);
> +		len = -EBUSY;
> +		goto out;
> +	}
> +
> +	if (ctx->counter_type == DAMOS_NUMBER_WRITES)
> +		len = scnprintf(kbuf, count, "writes");
> +	else
> +		len = scnprintf(kbuf, count, "accesses");
> +	mutex_unlock(&ctx->kdamond_lock);
> +	len = simple_read_from_buffer(buf, count, ppos, kbuf, len);
> +
> +out:
> +	kfree(kbuf);
> +	return len;
> +}
> +
> +
>  static int add_init_region(struct damon_ctx *c,
>  			 unsigned long target_id, struct damon_addr_range *ar)
>  {
> @@ -636,12 +698,18 @@ static const struct file_operations kdamond_pid_fops = {
>  	.read = dbgfs_kdamond_pid_read,
>  };
>  
> +static const struct file_operations counter_type_fops = {
> +	.open = damon_dbgfs_open,
> +	.read = dbgfs_counter_type_read,
> +	.write = dbgfs_counter_type_write,
> +};
> +
>  static void dbgfs_fill_ctx_dir(struct dentry *dir, struct damon_ctx *ctx)
>  {
>  	const char * const file_names[] = {"attrs", "schemes", "target_ids",
> -		"init_regions", "kdamond_pid"};
> +		"init_regions", "kdamond_pid", "counter_type"};
>  	const struct file_operations *fops[] = {&attrs_fops, &schemes_fops,
> -		&target_ids_fops, &init_regions_fops, &kdamond_pid_fops};
> +		&target_ids_fops, &init_regions_fops, &kdamond_pid_fops, &counter_type_fops};
>  	int i;
>  
>  	for (i = 0; i < ARRAY_SIZE(file_names); i++)
> diff --git a/mm/damon/prmtv-common.c b/mm/damon/prmtv-common.c
> index 92a04f5831d6..09ba2b5b895e 100644
> --- a/mm/damon/prmtv-common.c
> +++ b/mm/damon/prmtv-common.c
> @@ -9,6 +9,8 @@
>  #include <linux/page_idle.h>
>  #include <linux/pagemap.h>
>  #include <linux/rmap.h>
> +#include <linux/swap.h>
> +#include <linux/swapops.h>
>  
>  #include "prmtv-common.h"
>  
> @@ -58,6 +60,92 @@ void damon_ptep_mkold(pte_t *pte, struct mm_struct *mm, unsigned long addr)
>  	put_page(page);
>  }
>  
> +static inline bool pte_is_pinned(struct vm_area_struct *vma, unsigned long addr, pte_t pte)
> +{
> +	struct page *page;
> +
> +	if (!pte_write(pte))
> +		return false;
> +	if (!is_cow_mapping(vma->vm_flags))
> +		return false;
> +	if (likely(!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags)))
> +		return false;
> +	page = vm_normal_page(vma, addr, pte);
> +	if (!page)
> +		return false;
> +	return page_maybe_dma_pinned(page);
> +}
> +
> +static inline void clear_soft_dirty_pmd(struct vm_area_struct *vma,
> +		unsigned long addr, pmd_t *pmdp)
> +{
> +	pmd_t old, pmd = *pmdp;
> +
> +	if (pmd_present(pmd)) {
> +		/* See comment in change_huge_pmd() */
> +		old = pmdp_invalidate(vma, addr, pmdp);
> +		if (pmd_dirty(old))
> +			pmd = pmd_mkdirty(pmd);
> +		if (pmd_young(old))
> +			pmd = pmd_mkyoung(pmd);
> +
> +		pmd = pmd_wrprotect(pmd);
> +		pmd = pmd_clear_soft_dirty(pmd);
> +
> +		set_pmd_at(vma->vm_mm, addr, pmdp, pmd);
> +	} else if (is_migration_entry(pmd_to_swp_entry(pmd))) {
> +		pmd = pmd_swp_clear_soft_dirty(pmd);
> +		set_pmd_at(vma->vm_mm, addr, pmdp, pmd);
> +	}
> +}
> +
> +static inline void clear_soft_dirty(struct vm_area_struct *vma,
> +		unsigned long addr, pte_t *pte)
> +{
> +	/*
> +	 * The soft-dirty tracker uses #PF-s to catch writes
> +	 * to pages, so write-protect the pte as well. See the
> +	 * Documentation/admin-guide/mm/soft-dirty.rst for full description
> +	 * of how soft-dirty works.
> +	 */
> +	pte_t ptent = *pte;
> +
> +	if (pte_present(ptent)) {
> +		pte_t old_pte;
> +
> +		if (pte_is_pinned(vma, addr, ptent))
> +			return;
> +		old_pte = ptep_modify_prot_start(vma, addr, pte);
> +		ptent = pte_wrprotect(old_pte);
> +		ptent = pte_clear_soft_dirty(ptent);
> +		ptep_modify_prot_commit(vma, addr, pte, old_pte, ptent);
> +	} else if (is_swap_pte(ptent)) {
> +		ptent = pte_swp_clear_soft_dirty(ptent);
> +		set_pte_at(vma->vm_mm, addr, pte, ptent);
> +	}
> +}
> +
> +void damon_pmd_clean_soft_dirty(struct vm_area_struct *vma, unsigned long addr,
> +		pmd_t *pmd)
> +{
> +	vma->vm_flags &= ~VM_SOFTDIRTY;
> +	vma_set_page_prot(vma);
> +
> +	if (pmd_soft_dirty(*pmd))
> +		clear_soft_dirty_pmd(vma, addr, pmd);
> +
> +}
> +
> +void damon_ptep_clean_soft_dirty(struct vm_area_struct *vma, unsigned long addr,
> +		pte_t *pte)
> +{
> +	vma->vm_flags &= ~VM_SOFTDIRTY;
> +	vma_set_page_prot(vma);
> +
> +	if (pte_soft_dirty(*pte))
> +		clear_soft_dirty(vma, addr, pte);
> +}
> +
>  void damon_pmdp_mkold(pmd_t *pmd, struct mm_struct *mm, unsigned long addr)
>  {
>  #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> diff --git a/mm/damon/prmtv-common.h b/mm/damon/prmtv-common.h
> index 61f27037603e..03847d149f0e 100644
> --- a/mm/damon/prmtv-common.h
> +++ b/mm/damon/prmtv-common.h
> @@ -13,6 +13,11 @@
>  
>  struct page *damon_get_page(unsigned long pfn);
>  
> +void damon_ptep_clean_soft_dirty(struct vm_area_struct *vma, unsigned long addr,
> +		pte_t *pte);
> +void damon_pmd_clean_soft_dirty(struct vm_area_struct *vma, unsigned long addr,
> +		pmd_t *pmd);
> +
>  void damon_ptep_mkold(pte_t *pte, struct mm_struct *mm, unsigned long addr);
>  void damon_pmdp_mkold(pmd_t *pmd, struct mm_struct *mm, unsigned long addr);
>  
> diff --git a/mm/damon/vaddr.c b/mm/damon/vaddr.c
> index 20a9a9d69eb1..a7d9c9563d1d 100644
> --- a/mm/damon/vaddr.c
> +++ b/mm/damon/vaddr.c
> @@ -368,6 +368,44 @@ void damon_va_update(struct damon_ctx *ctx)
>  	}
>  }
>  
> +static int damon_clean_soft_dirty_pmd_entry(pmd_t *pmd, unsigned long addr,
> +		unsigned long next, struct mm_walk *walk)
> +{
> +	pte_t *pte;
> +	spinlock_t *ptl;
> +
> +	if (pmd_huge(*pmd)) {
> +		ptl = pmd_lock(walk->mm, pmd);
> +		if (pmd_huge(*pmd)) {
> +			damon_pmd_clean_soft_dirty(walk->vma, addr, pmd);
> +			spin_unlock(ptl);
> +			return 0;
> +		}
> +		spin_unlock(ptl);
> +	}
> +
> +	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
> +		return 0;
> +	pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
> +	if (!pte_present(*pte))
> +		goto out;
> +	damon_ptep_clean_soft_dirty(walk->vma, addr, pte);
> +out:
> +	pte_unmap_unlock(pte, ptl);
> +	return 0;
> +}
> +
> +static const struct mm_walk_ops damon_clean_soft_dirty_ops = {
> +	.pmd_entry = damon_clean_soft_dirty_pmd_entry,
> +};
> +
> +static void damon_va_clean_soft_dirty(struct mm_struct *mm, unsigned long addr)
> +{
> +	mmap_read_lock(mm);
> +	walk_page_range(mm, addr, addr + 1, &damon_clean_soft_dirty_ops, NULL);
> +	mmap_read_unlock(mm);
> +}
> +
>  static int damon_mkold_pmd_entry(pmd_t *pmd, unsigned long addr,
>  		unsigned long next, struct mm_walk *walk)
>  {
> @@ -415,7 +453,10 @@ static void damon_va_prepare_access_check(struct damon_ctx *ctx,
>  {
>  	r->sampling_addr = damon_rand(r->ar.start, r->ar.end);
>  
> -	damon_va_mkold(mm, r->sampling_addr);
> +	if (ctx->counter_type == DAMOS_NUMBER_WRITES)
> +		damon_va_clean_soft_dirty(mm, r->sampling_addr);
> +	else
> +		damon_va_mkold(mm, r->sampling_addr);
>  }
>  
>  void damon_va_prepare_access_checks(struct damon_ctx *ctx)
> @@ -437,6 +478,7 @@ void damon_va_prepare_access_checks(struct damon_ctx *ctx)
>  struct damon_young_walk_private {
>  	unsigned long *page_sz;
>  	bool young;
> +	bool dirty;
>  };
>  
>  static int damon_young_pmd_entry(pmd_t *pmd, unsigned long addr,
> @@ -463,6 +505,10 @@ static int damon_young_pmd_entry(pmd_t *pmd, unsigned long addr,
>  			*priv->page_sz = ((1UL) << HPAGE_PMD_SHIFT);
>  			priv->young = true;
>  		}
> +		if (pmd_soft_dirty(*pmd)) {
> +			*priv->page_sz = ((1UL) << HPAGE_PMD_SHIFT);
> +			priv->dirty = true;
> +		}
>  		put_page(page);
>  huge_out:
>  		spin_unlock(ptl);
> @@ -485,6 +531,10 @@ static int damon_young_pmd_entry(pmd_t *pmd, unsigned long addr,
>  		*priv->page_sz = PAGE_SIZE;
>  		priv->young = true;
>  	}
> +	if (pte_soft_dirty(*pte)) {
> +		*priv->page_sz = PAGE_SIZE;
> +		priv->dirty = true;
> +	}
>  	put_page(page);
>  out:
>  	pte_unmap_unlock(pte, ptl);
> @@ -496,16 +546,19 @@ static const struct mm_walk_ops damon_young_ops = {
>  };
>  
>  static bool damon_va_young(struct mm_struct *mm, unsigned long addr,
> -		unsigned long *page_sz)
> +		unsigned long *page_sz, bool *dirty)
>  {
>  	struct damon_young_walk_private arg = {
>  		.page_sz = page_sz,
>  		.young = false,
> +		.dirty = false,
>  	};
>  
>  	mmap_read_lock(mm);
>  	walk_page_range(mm, addr, addr + 1, &damon_young_ops, &arg);
>  	mmap_read_unlock(mm);
> +
> +	*dirty = arg.dirty;
>  	return arg.young;
>  }
>  
> @@ -522,18 +575,23 @@ static void damon_va_check_access(struct damon_ctx *ctx,
>  	static unsigned long last_addr;
>  	static unsigned long last_page_sz = PAGE_SIZE;
>  	static bool last_accessed;
> +	static bool last_written;
>  
>  	/* If the region is in the last checked page, reuse the result */
>  	if (mm == last_mm && (ALIGN_DOWN(last_addr, last_page_sz) ==
>  				ALIGN_DOWN(r->sampling_addr, last_page_sz))) {
>  		if (last_accessed)
>  			r->nr_accesses++;
> +		if (last_written)
> +			r->nr_writes++;
>  		return;
>  	}
>  
> -	last_accessed = damon_va_young(mm, r->sampling_addr, &last_page_sz);
> +	last_accessed = damon_va_young(mm, r->sampling_addr, &last_page_sz, &last_written);
>  	if (last_accessed)
>  		r->nr_accesses++;
> +	if (last_written)
> +		r->nr_writes++;
>  
>  	last_mm = mm;
>  	last_addr = r->sampling_addr;
> @@ -544,7 +602,7 @@ unsigned int damon_va_check_accesses(struct damon_ctx *ctx)
>  	struct damon_target *t;
>  	struct mm_struct *mm;
>  	struct damon_region *r;
> -	unsigned int max_nr_accesses = 0;
> +	unsigned int max_nr = 0;
>  
>  	damon_for_each_target(t, ctx) {
>  		mm = damon_get_mm(t);
> @@ -552,12 +610,15 @@ unsigned int damon_va_check_accesses(struct damon_ctx *ctx)
>  			continue;
>  		damon_for_each_region(r, t) {
>  			damon_va_check_access(ctx, mm, r);
> -			max_nr_accesses = max(r->nr_accesses, max_nr_accesses);
> +			if (ctx->counter_type == DAMOS_NUMBER_WRITES)
> +				max_nr = max(r->nr_writes, max_nr);
> +			else
> +				max_nr = max(r->nr_accesses, max_nr);
>  		}
>  		mmput(mm);
>  	}
>  
> -	return max_nr_accesses;
> +	return max_nr;
>  }
>  
>  /*
> @@ -597,6 +658,7 @@ static int damos_madvise(struct damon_target *target, struct damon_region *r,
>  
>  	ret = do_madvise(mm, PAGE_ALIGN(r->ar.start),
>  			PAGE_ALIGN(r->ar.end - r->ar.start), behavior);
> +
>  	mmput(mm);
>  out:
>  	return ret;
> @@ -624,6 +686,12 @@ int damon_va_apply_scheme(struct damon_ctx *ctx, struct damon_target *t,
>  	case DAMOS_NOHUGEPAGE:
>  		madv_action = MADV_NOHUGEPAGE;
>  		break;
> +	case DAMOS_MERGEABLE:
> +		madv_action = MADV_MERGEABLE;
> +		break;
> +	case DAMOS_UNMERGEABLE:
> +		madv_action = MADV_UNMERGEABLE;
> +		break;
>  	case DAMOS_STAT:
>  		return 0;
>  	default:
> -- 
> 2.25.1