[PATCH 1/1] mm: smaps: split PSS into components

[PATCH 1/1] mm: smaps: split PSS into components

[semenzato]

From: Luigi Semenzato <semenzato@chromium.org>

Report separate components (anon, file, and shmem)
for PSS in smaps_rollup.

This helps understand and tune the memory manager behavior
in consumer devices, particularly mobile devices.  Many of
them (e.g. chromebooks and Android-based devices) use zram
for anon memory, and perform disk reads for discarded file
pages.  The difference in latency is large (e.g. reading
a single page from SSD is 30 times slower than decompressing
a zram page on one popular device), thus it is useful to know
how much of the PSS is anon vs. file.

This patch also removes a small code duplication in smaps_account,
which would have gotten worse otherwise.

Also added missing entry for smaps_rollup in
Documentation/filesystems/proc.txt.

Acked-by: Yu Zhao <yuzhao@chromium.org>
Signed-off-by: Luigi Semenzato <semenzato@chromium.org>
---
 Documentation/filesystems/proc.txt |  6 +-
 fs/proc/task_mmu.c                 | 92 ++++++++++++++++++++----------
 2 files changed, 66 insertions(+), 32 deletions(-)

diff --git a/Documentation/filesystems/proc.txt b/Documentation/filesystems/proc.txt
index 66cad5c86171..b48e85e19877 100644
--- a/Documentation/filesystems/proc.txt
+++ b/Documentation/filesystems/proc.txt
@@ -153,9 +153,11 @@ Table 1-1: Process specific entries in /proc
 		symbol the task is blocked in - or "0" if not blocked.
  pagemap	Page table
  stack		Report full stack trace, enable via CONFIG_STACKTRACE
- smaps		an extension based on maps, showing the memory consumption of
+ smaps		An extension based on maps, showing the memory consumption of
 		each mapping and flags associated with it
- numa_maps	an extension based on maps, showing the memory locality and
+ smaps_rollup	Accumulated smaps stats for all mappings of the process.  This
+		can be derived from smaps, but is faster and more convenient
+ numa_maps	An extension based on maps, showing the memory locality and
 		binding policy as well as mem usage (in pages) of each mapping.
 ..............................................................................
 
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index 01d4eb0e6bd1..00d110dcd6c2 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -417,17 +417,53 @@ struct mem_size_stats {
 	unsigned long shared_hugetlb;
 	unsigned long private_hugetlb;
 	u64 pss;
+	u64 pss_anon;
+	u64 pss_file;
+	u64 pss_shmem;
 	u64 pss_locked;
 	u64 swap_pss;
 	bool check_shmem_swap;
 };
 
+static void smaps_page_accumulate(struct mem_size_stats *mss,
+		struct page *page, unsigned long size, unsigned long pss,
+		bool dirty, bool locked, bool private)
+{
+	mss->pss += pss;
+
+	if (PageAnon(page))
+		mss->pss_anon += pss;
+	else if (PageSwapBacked(page))
+		mss->pss_shmem += pss;
+	else
+		mss->pss_file += pss;
+
+	if (locked)
+		mss->pss_locked += pss;
+
+	if (dirty || PageDirty(page)) {
+		if (private)
+			mss->private_dirty += size;
+		else
+			mss->shared_dirty += size;
+	} else {
+		if (private)
+			mss->private_clean += size;
+		else
+			mss->shared_clean += size;
+	}
+}
+
 static void smaps_account(struct mem_size_stats *mss, struct page *page,
 		bool compound, bool young, bool dirty, bool locked)
 {
 	int i, nr = compound ? 1 << compound_order(page) : 1;
 	unsigned long size = nr * PAGE_SIZE;
 
+	/*
+	 * First accumulate quantities that depend only on |size| and the type
+	 * of the compound page.
+	 */
 	if (PageAnon(page)) {
 		mss->anonymous += size;
 		if (!PageSwapBacked(page) && !dirty && !PageDirty(page))
@@ -440,42 +476,25 @@ static void smaps_account(struct mem_size_stats *mss, struct page *page,
 		mss->referenced += size;
 
 	/*
+	 * Then accumulate quantities that may depend on sharing, or that may
+	 * differ page-by-page.
+	 *
 	 * page_count(page) == 1 guarantees the page is mapped exactly once.
 	 * If any subpage of the compound page mapped with PTE it would elevate
 	 * page_count().
 	 */
 	if (page_count(page) == 1) {
-		if (dirty || PageDirty(page))
-			mss->private_dirty += size;
-		else
-			mss->private_clean += size;
-		mss->pss += (u64)size << PSS_SHIFT;
-		if (locked)
-			mss->pss_locked += (u64)size << PSS_SHIFT;
+		smaps_page_accumulate(mss, page, size, size << PSS_SHIFT, dirty,
+			locked, true);
 		return;
 	}
-
 	for (i = 0; i < nr; i++, page++) {
 		int mapcount = page_mapcount(page);
-		unsigned long pss = (PAGE_SIZE << PSS_SHIFT);
-
-		if (mapcount >= 2) {
-			if (dirty || PageDirty(page))
-				mss->shared_dirty += PAGE_SIZE;
-			else
-				mss->shared_clean += PAGE_SIZE;
-			mss->pss += pss / mapcount;
-			if (locked)
-				mss->pss_locked += pss / mapcount;
-		} else {
-			if (dirty || PageDirty(page))
-				mss->private_dirty += PAGE_SIZE;
-			else
-				mss->private_clean += PAGE_SIZE;
-			mss->pss += pss;
-			if (locked)
-				mss->pss_locked += pss;
-		}
+		unsigned long pss = PAGE_SIZE << PSS_SHIFT;
+		if (mapcount >= 2)
+			pss /= mapcount;
+		smaps_page_accumulate(mss, page, PAGE_SIZE, pss, dirty, locked,
+				      mapcount < 2);
 	}
 }
 
@@ -754,10 +773,23 @@ static void smap_gather_stats(struct vm_area_struct *vma,
 		seq_put_decimal_ull_width(m, str, (val) >> 10, 8)
 
 /* Show the contents common for smaps and smaps_rollup */
-static void __show_smap(struct seq_file *m, const struct mem_size_stats *mss)
+static void __show_smap(struct seq_file *m, const struct mem_size_stats *mss,
+	bool rollup_mode)
 {
 	SEQ_PUT_DEC("Rss:            ", mss->resident);
 	SEQ_PUT_DEC(" kB\nPss:            ", mss->pss >> PSS_SHIFT);
+	if (rollup_mode) {
+		/*
+		 * These are meaningful only for smaps_rollup, otherwise two of
+		 * them are zero, and the other one is the same as Pss.
+		 */
+		SEQ_PUT_DEC(" kB\nPss_Anon:       ",
+			mss->pss_anon >> PSS_SHIFT);
+		SEQ_PUT_DEC(" kB\nPss_File:       ",
+			mss->pss_file >> PSS_SHIFT);
+		SEQ_PUT_DEC(" kB\nPss_Shmem:      ",
+			mss->pss_shmem >> PSS_SHIFT);
+	}
 	SEQ_PUT_DEC(" kB\nShared_Clean:   ", mss->shared_clean);
 	SEQ_PUT_DEC(" kB\nShared_Dirty:   ", mss->shared_dirty);
 	SEQ_PUT_DEC(" kB\nPrivate_Clean:  ", mss->private_clean);
@@ -794,7 +826,7 @@ static int show_smap(struct seq_file *m, void *v)
 	SEQ_PUT_DEC(" kB\nMMUPageSize:    ", vma_mmu_pagesize(vma));
 	seq_puts(m, " kB\n");
 
-	__show_smap(m, &mss);
+	__show_smap(m, &mss, false);
 
 	seq_printf(m, "THPeligible:    %d\n", transparent_hugepage_enabled(vma));
 
@@ -841,7 +873,7 @@ static int show_smaps_rollup(struct seq_file *m, void *v)
 	seq_pad(m, ' ');
 	seq_puts(m, "[rollup]\n");
 
-	__show_smap(m, &mss);
+	__show_smap(m, &mss, true);
 
 	release_task_mempolicy(priv);
 	up_read(&mm->mmap_sem);
-- 
2.22.0.410.gd8fdbe21b5-goog

Re: [PATCH 1/1] mm: smaps: split PSS into components

[Andrew Morton]

On Wed, 26 Jun 2019 11:04:29 -0700 semenzato@chromium.org wrote:

> From: Luigi Semenzato <semenzato@chromium.org>
> 
> Report separate components (anon, file, and shmem)
> for PSS in smaps_rollup.
> 
> This helps understand and tune the memory manager behavior
> in consumer devices, particularly mobile devices.  Many of
> them (e.g. chromebooks and Android-based devices) use zram
> for anon memory, and perform disk reads for discarded file
> pages.  The difference in latency is large (e.g. reading
> a single page from SSD is 30 times slower than decompressing
> a zram page on one popular device), thus it is useful to know
> how much of the PSS is anon vs. file.
> 
> This patch also removes a small code duplication in smaps_account,
> which would have gotten worse otherwise.
> 
> Also added missing entry for smaps_rollup in
> Documentation/filesystems/proc.txt.
> 
> ...
>
> -static void __show_smap(struct seq_file *m, const struct mem_size_stats *mss)
> +static void __show_smap(struct seq_file *m, const struct mem_size_stats *mss,
> +	bool rollup_mode)
>  {
>  	SEQ_PUT_DEC("Rss:            ", mss->resident);
>  	SEQ_PUT_DEC(" kB\nPss:            ", mss->pss >> PSS_SHIFT);
> +	if (rollup_mode) {
> +		/*
> +		 * These are meaningful only for smaps_rollup, otherwise two of
> +		 * them are zero, and the other one is the same as Pss.
> +		 */
> +		SEQ_PUT_DEC(" kB\nPss_Anon:       ",
> +			mss->pss_anon >> PSS_SHIFT);
> +		SEQ_PUT_DEC(" kB\nPss_File:       ",
> +			mss->pss_file >> PSS_SHIFT);
> +		SEQ_PUT_DEC(" kB\nPss_Shmem:      ",
> +			mss->pss_shmem >> PSS_SHIFT);
> +	}

Documentation/filesystems/proc.txt is rather incomplete.  It documents
/proc/PID/smaps (seems to be out of date) but doesn't describe the
fields in smaps_rollup.

Please update Documentation/ABI/testing/procfs-smaps_rollup and please
check that it's up-to-date while you're in there.

Re: [PATCH 1/1] mm: smaps: split PSS into components

[Luigi Semenzato]

On Wed, Jun 26, 2019 at 3:19 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>
> On Wed, 26 Jun 2019 11:04:29 -0700 semenzato@chromium.org wrote:
>
> > From: Luigi Semenzato <semenzato@chromium.org>
> >
> > Report separate components (anon, file, and shmem)
> > for PSS in smaps_rollup.
> >
> > This helps understand and tune the memory manager behavior
> > in consumer devices, particularly mobile devices.  Many of
> > them (e.g. chromebooks and Android-based devices) use zram
> > for anon memory, and perform disk reads for discarded file
> > pages.  The difference in latency is large (e.g. reading
> > a single page from SSD is 30 times slower than decompressing
> > a zram page on one popular device), thus it is useful to know
> > how much of the PSS is anon vs. file.
> >
> > This patch also removes a small code duplication in smaps_account,
> > which would have gotten worse otherwise.
> >
> > Also added missing entry for smaps_rollup in
> > Documentation/filesystems/proc.txt.
> >
> > ...
> >
> > -static void __show_smap(struct seq_file *m, const struct mem_size_stats *mss)
> > +static void __show_smap(struct seq_file *m, const struct mem_size_stats *mss,
> > +     bool rollup_mode)
> >  {
> >       SEQ_PUT_DEC("Rss:            ", mss->resident);
> >       SEQ_PUT_DEC(" kB\nPss:            ", mss->pss >> PSS_SHIFT);
> > +     if (rollup_mode) {
> > +             /*
> > +              * These are meaningful only for smaps_rollup, otherwise two of
> > +              * them are zero, and the other one is the same as Pss.
> > +              */
> > +             SEQ_PUT_DEC(" kB\nPss_Anon:       ",
> > +                     mss->pss_anon >> PSS_SHIFT);
> > +             SEQ_PUT_DEC(" kB\nPss_File:       ",
> > +                     mss->pss_file >> PSS_SHIFT);
> > +             SEQ_PUT_DEC(" kB\nPss_Shmem:      ",
> > +                     mss->pss_shmem >> PSS_SHIFT);
> > +     }
>
> Documentation/filesystems/proc.txt is rather incomplete.  It documents
> /proc/PID/smaps (seems to be out of date) but doesn't describe the
> fields in smaps_rollup.
>
> Please update Documentation/ABI/testing/procfs-smaps_rollup and please
> check that it's up-to-date while you're in there.
>

Thank you for noticing the stale/missing docs and sorry that I did not.
Will email the updated patch shortly.

[PATCH 1/1] mm: smaps: split PSS into components

[semenzato]

From: Luigi Semenzato <semenzato@chromium.org>

Report separate components (anon, file, and shmem)
for PSS in smaps and smaps_rollup.

This helps understand and tune the memory manager behavior
in consumer devices, particularly mobile devices.  Many of
them (e.g. chromebooks and Android-based devices) use zram
for anon memory, and perform disk reads for discarded file
pages.  The difference in latency is large (e.g. reading
a single page from SSD is 30 times slower than decompressing
a zram page on one popular device), thus it is useful to know
how much of the PSS is anon vs. file.

This patch also removes a small code duplication in smaps_account,
which would have gotten worse otherwise.

Signed-off-by: Luigi Semenzato <semenzato@chromium.org>
---
 fs/proc/task_mmu.c | 61 ++++++++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 24 deletions(-)

diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index 01d4eb0e6bd1..4b586c4d27b0 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -417,17 +417,45 @@ struct mem_size_stats {
 	unsigned long shared_hugetlb;
 	unsigned long private_hugetlb;
 	u64 pss;
+	u64 pss_anon;
+	u64 pss_file;
+	u64 pss_shmem;
 	u64 pss_locked;
 	u64 swap_pss;
 	bool check_shmem_swap;
 };
 
+static void smaps_page_accumulate(struct mem_size_stats *mss,
+		struct page *page, unsigned long size, unsigned long pss,
+		bool dirty, bool locked)
+{
+	mss->pss += pss;
+
+	if (PageAnon(page))
+		mss->pss_anon += pss;
+	else if (PageSwapBacked(page))
+		mss->pss_shmem += pss;
+	else
+		mss->pss_file += pss;
+
+	if (locked)
+		mss->pss_locked += pss;
+	if (dirty || PageDirty(page))
+		mss->shared_dirty += size;
+	else
+		mss->shared_clean += size;
+}
+
 static void smaps_account(struct mem_size_stats *mss, struct page *page,
 		bool compound, bool young, bool dirty, bool locked)
 {
 	int i, nr = compound ? 1 << compound_order(page) : 1;
 	unsigned long size = nr * PAGE_SIZE;
 
+	/*
+	 * First accumulate quantities that depend only on |size| and the type
+	 * of the compound page.
+	 */
 	if (PageAnon(page)) {
 		mss->anonymous += size;
 		if (!PageSwapBacked(page) && !dirty && !PageDirty(page))
@@ -440,42 +468,24 @@ static void smaps_account(struct mem_size_stats *mss, struct page *page,
 		mss->referenced += size;
 
 	/*
+	 * Then accumulate quantities that may depend on sharing, or that may
+	 * differ page-by-page.
+	 *
 	 * page_count(page) == 1 guarantees the page is mapped exactly once.
 	 * If any subpage of the compound page mapped with PTE it would elevate
 	 * page_count().
 	 */
 	if (page_count(page) == 1) {
-		if (dirty || PageDirty(page))
-			mss->private_dirty += size;
-		else
-			mss->private_clean += size;
-		mss->pss += (u64)size << PSS_SHIFT;
-		if (locked)
-			mss->pss_locked += (u64)size << PSS_SHIFT;
+		smaps_page_accumulate(mss, page, size, size, dirty, locked);
 		return;
 	}
-
 	for (i = 0; i < nr; i++, page++) {
 		int mapcount = page_mapcount(page);
 		unsigned long pss = (PAGE_SIZE << PSS_SHIFT);
-
 		if (mapcount >= 2) {
-			if (dirty || PageDirty(page))
-				mss->shared_dirty += PAGE_SIZE;
-			else
-				mss->shared_clean += PAGE_SIZE;
-			mss->pss += pss / mapcount;
-			if (locked)
-				mss->pss_locked += pss / mapcount;
-		} else {
-			if (dirty || PageDirty(page))
-				mss->private_dirty += PAGE_SIZE;
-			else
-				mss->private_clean += PAGE_SIZE;
-			mss->pss += pss;
-			if (locked)
-				mss->pss_locked += pss;
+			pss /= mapcount;
 		}
+		smaps_page_accumulate(mss, page, PAGE_SIZE, pss, dirty, locked);
 	}
 }
 
@@ -758,6 +768,9 @@ static void __show_smap(struct seq_file *m, const struct mem_size_stats *mss)
 {
 	SEQ_PUT_DEC("Rss:            ", mss->resident);
 	SEQ_PUT_DEC(" kB\nPss:            ", mss->pss >> PSS_SHIFT);
+	SEQ_PUT_DEC(" kB\nPss_Anon:       ", mss->pss_anon >> PSS_SHIFT);
+	SEQ_PUT_DEC(" kB\nPss_File:       ", mss->pss_file >> PSS_SHIFT);
+	SEQ_PUT_DEC(" kB\nPss_Shmem:      ", mss->pss_shmem >> PSS_SHIFT);
 	SEQ_PUT_DEC(" kB\nShared_Clean:   ", mss->shared_clean);
 	SEQ_PUT_DEC(" kB\nShared_Dirty:   ", mss->shared_dirty);
 	SEQ_PUT_DEC(" kB\nPrivate_Clean:  ", mss->private_clean);
-- 
2.21.0.1020.gf2820cf01a-goog