Skip to contentSkip to conversation

dax: Flush partial PMDs correctly

PATCH In review Matthew Wilcox 2 participants 3 messages Mar 1, 2019

Subsystems (3)
lore
Share & tools
lore

Matthew Wilcox

author reviewer
#0 lore raw

PATCH 1/1 dax: Flush partial PMDs correctly

The radix tree would rewind the index in an iterator to the lowest index of a multi-slot entry. The XArray iterators instead leave the index unchanged, but I overlooked that when converting DAX from the radix tree to the XArray. Adjust the index that we use for flushing to the start of the PMD range.

Fixes c1901cd33cf4 "page cache: Convert find_get_entries_tag to XArray"
---
 fs/dax.c | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)
fs/dax.c +8 -9
index 6959837cc465..f7a7af766efe 100644
@@ -843,9 +843,8 @@ static void dax_entry_mkclean(struct address_space *mapping, pgoff_t index,
 static int dax_writeback_one(struct xa_state *xas, struct dax_device *dax_dev,
 		struct address_space *mapping, void *entry)
 {
-	unsigned long pfn;
+	unsigned long pfn, index, count;
 	long ret = 0;
-	size_t size;
 
 	/*
 	 * A page got tagged dirty in DAX mapping? Something is seriously
@@ -894,17 +893,18 @@ static int dax_writeback_one(struct xa_state *xas, struct dax_device *dax_dev,
 	xas_unlock_irq(xas);
 
 	/*
-	 * Even if dax_writeback_mapping_range() was given a wbc->range_start
-	 * in the middle of a PMD, the 'index' we are given will be aligned to
-	 * the start index of the PMD, as will the pfn we pull from 'entry'.
+	 * If dax_writeback_mapping_range() was given a wbc->range_start
+	 * in the middle of a PMD, the 'index' we are given needs to be
+	 * aligned to the start index of the PMD.
 	 * This allows us to flush for PMD_SIZE and not have to worry about
 	 * partial PMD writebacks.
 	 */
 	pfn = dax_to_pfn(entry);
-	size = PAGE_SIZE << dax_entry_order(entry);
+	count = 1UL << dax_entry_order(entry);
+	index = xas->xa_index &~ (count - 1);
 
 	dax_entry_mkclean(mapping, xas->xa_index, pfn);
-	dax_flush(dax_dev, page_address(pfn_to_page(pfn)), size);
+	dax_flush(dax_dev, page_address(pfn_to_page(pfn)), count * PAGE_SIZE);
 	/*
 	 * After we have flushed the cache, we can clear the dirty tag. There
 	 * cannot be new dirty data in the pfn after the flush has completed as
@@ -917,8 +917,7 @@ static int dax_writeback_one(struct xa_state *xas, struct dax_device *dax_dev,
 	xas_clear_mark(xas, PAGECACHE_TAG_DIRTY);
 	dax_wake_entry(xas, entry, false);
 
-	trace_dax_writeback_one(mapping->host, xas->xa_index,
-			size >> PAGE_SHIFT);
+	trace_dax_writeback_one(mapping->host, xas->xa_index, count);
 	return ret;
 
  put_unlocked:
Signature · 1 line
2.20.1
reviewer
#1 lore raw
Quoted 11 lines · Matthew Wilcox

The radix tree would rewind the index in an iterator to the lowest index of a multi-slot entry. The XArray iterators instead leave the index unchanged, but I overlooked that when converting DAX from the radix tree to the XArray. Adjust the index that we use for flushing to the start of the PMD range.

Fixes c1901cd33cf4 "page cache: Convert find_get_entries_tag to XArray"

---

Thanks for the patch! One comment below:

Quoted 7 lines
fs/dax.c +0 -3
@@ -894,17 +893,18 @@ static int dax_writeback_one(struct xa_state *xas, struct dax_device *dax_dev,
 	xas_unlock_irq(xas);
 
 	/*
-	 * Even if dax_writeback_mapping_range() was given a wbc->range_start
-	 * in the middle of a PMD, the 'index' we are given will be aligned to
-	 * the start index of the PMD, as will the pfn we pull from 'entry'.

fs/dax.c static int dax_writeback_one(struct xa_state *xas, struct dax_device *dax_dev, · line 904

Quoted 10 lines
fs/dax.c +5 -1
+	 * If dax_writeback_mapping_range() was given a wbc->range_start
+	 * in the middle of a PMD, the 'index' we are given needs to be
+	 * aligned to the start index of the PMD.
 	 * This allows us to flush for PMD_SIZE and not have to worry about
 	 * partial PMD writebacks.
 	 */
 	pfn = dax_to_pfn(entry);
-	size = PAGE_SIZE << dax_entry_order(entry);
+	count = 1UL << dax_entry_order(entry);
+	index = xas->xa_index &~ (count - 1);

Hum, why do you compute 'index' here when you actually never use it? The whole patch looks fishy since it is effectively a noop AFAICT...

Honza

Quoted 20 lines
fs/dax.c +2 -3
 	dax_entry_mkclean(mapping, xas->xa_index, pfn);
-	dax_flush(dax_dev, page_address(pfn_to_page(pfn)), size);
+	dax_flush(dax_dev, page_address(pfn_to_page(pfn)), count * PAGE_SIZE);
 	/*
 	 * After we have flushed the cache, we can clear the dirty tag. There
 	 * cannot be new dirty data in the pfn after the flush has completed as
@@ -917,8 +917,7 @@ static int dax_writeback_one(struct xa_state *xas, struct dax_device *dax_dev,
 	xas_clear_mark(xas, PAGECACHE_TAG_DIRTY);
 	dax_wake_entry(xas, entry, false);
 
-	trace_dax_writeback_one(mapping->host, xas->xa_index,
-			size >> PAGE_SHIFT);
+	trace_dax_writeback_one(mapping->host, xas->xa_index, count);
 	return ret;
 
  put_unlocked:
Signature · 1 line
2.20.1
Signature · 6 lines
Jan Kara <[email protected]>
SUSE Labs, CR
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
author reviewer
#2 lore raw
Quoted 7 lines · Jan Kara

fs/dax.c static int dax_writeback_one(struct xa_state *xas, struct dax_device *dax_dev, · line 904

Quoted 4 lines
fs/dax.c +2 -1
 	pfn = dax_to_pfn(entry);
-	size = PAGE_SIZE << dax_entry_order(entry);
+	count = 1UL << dax_entry_order(entry);
+	index = xas->xa_index &~ (count - 1);

Hum, why do you compute 'index' here when you actually never use it? The whole patch looks fishy since it is effectively a noop AFAICT...

*facepalm*. I lost the change to the next line while I was backing out some extraneous changes.

Quoted 1 line
Quoted 1 line

dax_entry_mkclean(mapping, xas->xa_index, pfn);

Replacement patch sent.
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm