Skip to contentSkip to conversation

mm: optimize shadow entries removal

PATCH Superseded Shakeel Butt 2 participants 5 messages Sep 11, 2024

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover mm: optimize shadow entries removal

Some of our production workloads which processes a large amount of data spends considerable amount of CPUs on truncation and invalidation of large sized files (100s of GiBs of size). Tracing the operations showed that most of the time is in shadow entries removal. This patch series optimizes the truncation and invalidation operations.

(This is 6.13+ material)

Shakeel Butt (2):
mm: optimize truncation of shadow entries
mm: optimize invalidation of shadow entries

 mm/truncate.c | 96 ++++++++++++++++++++++-----------------------------
 1 file changed, 41 insertions(+), 55 deletions(-)
Signature · 1 line
2.43.5
#1 lore raw

PATCH 1/2 mm: optimize truncation of shadow entries

The kernel truncates the page cache in batches of PAGEVEC_SIZE. For each batch, it traverses the page cache tree and collects the entries (folio and shadow entries) in the struct folio_batch. For the shadow entries present in the folio_batch, it has to traverse the page cache tree for each individual entry to remove them. This patch optimize this by removing them in a single tree traversal.

On large machines in our production which run workloads manipulating large amount of data, we have observed that a large amount of CPUs are spent on truncation of very large files (100s of GiBs file sizes). More specifically most of time was spent on shadow entries cleanup, so optimizing the shadow entries cleanup, even a little bit, has good impact.

To evaluate the changes, we created 200GiB file on a fuse fs and in a memcg. We created the shadow entries by triggering reclaim through memory.reclaim in that specific memcg and measure the simple truncation operation.

 # time truncate -s 0 file

time (sec)
Without 5.164 +- 0.059
With-patch 4.21 +- 0.066 (18.47% decrease)

---
 mm/truncate.c | 50 +++++++++++++++++++++++---------------------------
 1 file changed, 23 insertions(+), 27 deletions(-)
mm/truncate.c +23 -27
index 0668cd340a46..c7c19c816c2e 100644
@@ -72,50 +72,46 @@ static void clear_shadow_entries(struct address_space *mapping,
 static void truncate_folio_batch_exceptionals(struct address_space *mapping,
 				struct folio_batch *fbatch, pgoff_t *indices)
 {
+	XA_STATE(xas, &mapping->i_pages, indices[0]);
+	int nr = folio_batch_count(fbatch);
+	struct folio *folio;
 	int i, j;
-	bool dax;
 
 	/* Handled by shmem itself */
 	if (shmem_mapping(mapping))
 		return;
 
-	for (j = 0; j < folio_batch_count(fbatch); j++)
+	for (j = 0; j < nr; j++)
 		if (xa_is_value(fbatch->folios[j]))
 			break;
 
-	if (j == folio_batch_count(fbatch))
+	if (j == nr)
 		return;
 
-	dax = dax_mapping(mapping);
-	if (!dax) {
-		spin_lock(&mapping->host->i_lock);
-		xa_lock_irq(&mapping->i_pages);
+	if (dax_mapping(mapping)) {
+		for (i = j; i < nr; i++) {
+			if (xa_is_value(fbatch->folios[i]))
+				dax_delete_mapping_entry(mapping, indices[i]);
+		}
+		goto out;
 	}
 
-	for (i = j; i < folio_batch_count(fbatch); i++) {
-		struct folio *folio = fbatch->folios[i];
-		pgoff_t index = indices[i];
-
-		if (!xa_is_value(folio)) {
-			fbatch->folios[j++] = folio;
-			continue;
-		}
+	xas_set_update(&xas, workingset_update_node);
 
-		if (unlikely(dax)) {
-			dax_delete_mapping_entry(mapping, index);
-			continue;
-		}
+	spin_lock(&mapping->host->i_lock);
+	xas_lock_irq(&xas);
 
-		__clear_shadow_entry(mapping, index, folio);
+	xas_for_each(&xas, folio, indices[nr-1]) {
+		if (xa_is_value(folio))
+			xas_store(&xas, NULL);
 	}
 
-	if (!dax) {
-		xa_unlock_irq(&mapping->i_pages);
-		if (mapping_shrinkable(mapping))
-			inode_add_lru(mapping->host);
-		spin_unlock(&mapping->host->i_lock);
-	}
-	fbatch->nr = j;
+	xas_unlock_irq(&xas);
+	if (mapping_shrinkable(mapping))
+		inode_add_lru(mapping->host);
+	spin_unlock(&mapping->host->i_lock);
+out:
+	folio_batch_remove_exceptionals(fbatch);
 }
 
 /**
Signature · 1 line
2.43.5
#4 lore raw

PATCH 2/2 mm: optimize invalidation of shadow entries

In reply to Shakeel Butt

The kernel invalidates the page cache in batches of PAGEVEC_SIZE. For each batch, it traverses the page cache tree and collects the entries (folio and shadow entries) in the struct folio_batch. For the shadow entries present in the folio_batch, it has to traverse the page cache tree for each individual entry to remove them. This patch optimize this by removing them in a single tree traversal.

To evaluate the changes, we created 200GiB file on a fuse fs and in a memcg. We created the shadow entries by triggering reclaim through memory.reclaim in that specific memcg and measure the simple fadvise(DONTNEED) operation.

 # time xfs_io -c 'fadvise -d 0 ${file_size}' file

time (sec)
Without 5.12 +- 0.061
With-patch 4.19 +- 0.086 (18.16% decrease)

---
 mm/truncate.c | 46 ++++++++++++++++++----------------------------
 1 file changed, 18 insertions(+), 28 deletions(-)
mm/truncate.c +18 -28
index c7c19c816c2e..793c0d17d7b4 100644
@@ -23,42 +23,28 @@
 #include <linux/rmap.h>
 #include "internal.h"
 
-/*
- * Regular page slots are stabilized by the page lock even without the tree
- * itself locked.  These unlocked entries need verification under the tree
- * lock.
- */
-static inline void __clear_shadow_entry(struct address_space *mapping,
-				pgoff_t index, void *entry)
-{
-	XA_STATE(xas, &mapping->i_pages, index);
-
-	xas_set_update(&xas, workingset_update_node);
-	if (xas_load(&xas) != entry)
-		return;
-	xas_store(&xas, NULL);
-}
-
 static void clear_shadow_entries(struct address_space *mapping,
-				 struct folio_batch *fbatch, pgoff_t *indices)
+				 unsigned long start, unsigned long max)
 {
-	int i;
+	XA_STATE(xas, &mapping->i_pages, start);
+	struct folio *folio;
 
 	/* Handled by shmem itself, or for DAX we do nothing. */
 	if (shmem_mapping(mapping) || dax_mapping(mapping))
 		return;
 
-	spin_lock(&mapping->host->i_lock);
-	xa_lock_irq(&mapping->i_pages);
+	xas_set_update(&xas, workingset_update_node);
 
-	for (i = 0; i < folio_batch_count(fbatch); i++) {
-		struct folio *folio = fbatch->folios[i];
+	spin_lock(&mapping->host->i_lock);
+	xas_lock_irq(&xas);
 
+	/* Clear all shadow entries from start to max */
+	xas_for_each(&xas, folio, max) {
 		if (xa_is_value(folio))
-			__clear_shadow_entry(mapping, indices[i], folio);
+			xas_store(&xas, NULL);
 	}
 
-	xa_unlock_irq(&mapping->i_pages);
+	xas_unlock_irq(&xas);
 	if (mapping_shrinkable(mapping))
 		inode_add_lru(mapping->host);
 	spin_unlock(&mapping->host->i_lock);
@@ -478,7 +464,9 @@ unsigned long mapping_try_invalidate(struct address_space *mapping,
 
 	folio_batch_init(&fbatch);
 	while (find_lock_entries(mapping, &index, end, &fbatch, indices)) {
-		for (i = 0; i < folio_batch_count(&fbatch); i++) {
+		int nr = folio_batch_count(&fbatch);
+
+		for (i = 0; i < nr; i++) {
 			struct folio *folio = fbatch.folios[i];
 
 			/* We rely upon deletion not changing folio->index */
@@ -505,7 +493,7 @@ unsigned long mapping_try_invalidate(struct address_space *mapping,
 		}
 
 		if (xa_has_values)
-			clear_shadow_entries(mapping, &fbatch, indices);
+			clear_shadow_entries(mapping, indices[0], indices[nr-1]);
 
 		folio_batch_remove_exceptionals(&fbatch);
 		folio_batch_release(&fbatch);
@@ -609,7 +597,9 @@ int invalidate_inode_pages2_range(struct address_space *mapping,
 	folio_batch_init(&fbatch);
 	index = start;
 	while (find_get_entries(mapping, &index, end, &fbatch, indices)) {
-		for (i = 0; i < folio_batch_count(&fbatch); i++) {
+		int nr = folio_batch_count(&fbatch);
+
+		for (i = 0; i < nr; i++) {
 			struct folio *folio = fbatch.folios[i];
 
 			/* We rely upon deletion not changing folio->index */
@@ -655,7 +645,7 @@ int invalidate_inode_pages2_range(struct address_space *mapping,
 		}
 
 		if (xa_has_values)
-			clear_shadow_entries(mapping, &fbatch, indices);
+			clear_shadow_entries(mapping, indices[0], indices[nr-1]);
 
 		folio_batch_remove_exceptionals(&fbatch);
 		folio_batch_release(&fbatch);
Signature · 1 line
2.43.5
Acked-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] mm: optimize truncation of shadow entries

In reply to Shakeel Butt
Quoted 19 lines · Shakeel Butt

The kernel truncates the page cache in batches of PAGEVEC_SIZE. For each batch, it traverses the page cache tree and collects the entries (folio and shadow entries) in the struct folio_batch. For the shadow entries present in the folio_batch, it has to traverse the page cache tree for each individual entry to remove them. This patch optimize this by removing them in a single tree traversal.

On large machines in our production which run workloads manipulating large amount of data, we have observed that a large amount of CPUs are spent on truncation of very large files (100s of GiBs file sizes). More specifically most of time was spent on shadow entries cleanup, so optimizing the shadow entries cleanup, even a little bit, has good impact.

To evaluate the changes, we created 200GiB file on a fuse fs and in a memcg. We created the shadow entries by triggering reclaim through memory.reclaim in that specific memcg and measure the simple truncation operation.

Quoted 7 lines
 # time truncate -s 0 file

time (sec)
Without 5.164 +- 0.059
With-patch 4.21 +- 0.066 (18.47% decrease)

Looks good to me. One thing that's a bit subtle is that the tree walk assumes indices[] are ordered, such that indices[0] and indices[nr-1] reliably denote the range of interest. AFAICS that's the case for the current callers but if not that could be a painful bug to hunt down.

Assessing lowest and highest index in that first batch iteration seems a bit overkill though. Maybe just a comment stating the requirement?

Otherwise,

#3 lore raw

Re: [PATCH 1/2] mm: optimize truncation of shadow entries

Quoted 26 lines · Johannes Weiner
Quoted 25 lines · Shakeel Butt

The kernel truncates the page cache in batches of PAGEVEC_SIZE. For each batch, it traverses the page cache tree and collects the entries (folio and shadow entries) in the struct folio_batch. For the shadow entries present in the folio_batch, it has to traverse the page cache tree for each individual entry to remove them. This patch optimize this by removing them in a single tree traversal.

On large machines in our production which run workloads manipulating large amount of data, we have observed that a large amount of CPUs are spent on truncation of very large files (100s of GiBs file sizes). More specifically most of time was spent on shadow entries cleanup, so optimizing the shadow entries cleanup, even a little bit, has good impact.

To evaluate the changes, we created 200GiB file on a fuse fs and in a memcg. We created the shadow entries by triggering reclaim through memory.reclaim in that specific memcg and measure the simple truncation operation.

 # time truncate -s 0 file

time (sec)
Without 5.164 +- 0.059
With-patch 4.21 +- 0.066 (18.47% decrease)

Quoted 6 lines
Quoted 1 line

Looks good to me. One thing that's a bit subtle is that the tree walk assumes indices[] are ordered, such that indices[0] and indices[nr-1] reliably denote the range of interest. AFAICS that's the case for the current callers but if not that could be a painful bug to hunt down.

The current callers use find_get_entries() and find_lock_entries() to fill up the indices array which provides this guarantee.

Quoted 3 lines

Assessing lowest and highest index in that first batch iteration seems a bit overkill though. Maybe just a comment stating the requirement?

I will add a comment in v2.

Quoted 4 lines

Otherwise,

Thanks for the review.