[v3] mm: Convert DAX lock/unlock page to lock/unlock folio

Commit Message

Jane Chu Sept. 8, 2023, 10:23 p.m. UTC

From Matthew Wilcox:

The one caller of DAX lock/unlock page already calls compound_head(),
so use page_folio() instead, then use a folio throughout the DAX code
to remove uses of page->mapping and page->index. [1]

The additional change to [1] is comments added to mf_generic_kill_procs().

[1] https://lore.kernel.org/linux-mm/[e-mail]/T/

Signed-off-by: Matthew Wilcox (Oracle) <[e-mail]>
Acked-by: Naoya Horiguchi <[e-mail]>
Signed-off-by: Jane Chu <[e-mail]>
---
 fs/dax.c            | 24 ++++++++++++------------
 include/linux/dax.h | 10 +++++-----
 mm/memory-failure.c | 29 ++++++++++++++++-------------
 3 files changed, 33 insertions(+), 30 deletions(-)


base-commit: 727dbda16b83600379061c4ca8270ef3e2f51922

Comments

Andrew Morton Sept. 10, 2023, 7:46 p.m. UTC | #1

On Fri,  8 Sep 2023 16:23:36 -0600 Jane Chu <[e-mail]> wrote:

> >From Matthew Wilcox:
> 
> The one caller of DAX lock/unlock page already calls compound_head(),
> so use page_folio() instead, then use a folio throughout the DAX code
> to remove uses of page->mapping and page->index. [1]
> 
> The additional change to [1] is comments added to mf_generic_kill_procs().
> 
> [1] https://lore.kernel.org/linux-mm/[e-mail]/T/
> 

The delta versus the patch which is presently in mm.git is:

--- a/mm/memory-failure.c~a
+++ a/mm/memory-failure.c
@@ -1720,11 +1720,19 @@ static void unmap_and_kill(struct list_h
 	kill_procs(to_kill, flags & MF_MUST_KILL, false, pfn, flags);
 }
 
+/*
+ * Only dev_pagemap pages get here, such as fsdax when the filesystem
+ * either do not claim or fails to claim a hwpoison event, or devdax.
+ * The fsdax pages are initialized per base page, and the devdax pages
+ * could be initialized either as base pages, or as compound pages with
+ * vmemmap optimization enabled. Devdax is simplistic in its dealing with
+ * hwpoison, such that, if a subpage of a compound page is poisoned,
+ * simply mark the compound head page is by far sufficient.
+ */
 static int mf_generic_kill_procs(unsigned long long pfn, int flags,
 		struct dev_pagemap *pgmap)
 {
-	struct page *page = pfn_to_page(pfn);
-	struct folio *folio = page_folio(page);
+	struct folio *folio = pfn_folio(pfn);
 	LIST_HEAD(to_kill);
 	dax_entry_t cookie;
 	int rc = 0;

so I assume this is the v1->v3 delta.

I'll queue this as a fixup patch with the changelog

add comment to mf_generic_kill_procss(), simplify
mf_generic_kill_procs:folio initialization.

Jane Chu Sept. 11, 2023, 4:37 p.m. UTC | #2

Hi, Andrew,

On 9/10/2023 12:46 PM, Andrew Morton wrote:
> On Fri,  8 Sep 2023 16:23:36 -0600 Jane Chu <[e-mail]> wrote:
> 
>> >From Matthew Wilcox:
>>
>> The one caller of DAX lock/unlock page already calls compound_head(),
>> so use page_folio() instead, then use a folio throughout the DAX code
>> to remove uses of page->mapping and page->index. [1]
>>
>> The additional change to [1] is comments added to mf_generic_kill_procs().
>>
>> [1] https://lore.kernel.org/linux-mm/[e-mail]/T/
>>
> 
> The delta versus the patch which is presently in mm.git is:
> 
> --- a/mm/memory-failure.c~a
> +++ a/mm/memory-failure.c
> @@ -1720,11 +1720,19 @@ static void unmap_and_kill(struct list_h
>   	kill_procs(to_kill, flags & MF_MUST_KILL, false, pfn, flags);
>   }
>   
> +/*
> + * Only dev_pagemap pages get here, such as fsdax when the filesystem
> + * either do not claim or fails to claim a hwpoison event, or devdax.
> + * The fsdax pages are initialized per base page, and the devdax pages
> + * could be initialized either as base pages, or as compound pages with
> + * vmemmap optimization enabled. Devdax is simplistic in its dealing with
> + * hwpoison, such that, if a subpage of a compound page is poisoned,
> + * simply mark the compound head page is by far sufficient.
> + */
>   static int mf_generic_kill_procs(unsigned long long pfn, int flags,
>   		struct dev_pagemap *pgmap)
>   {
> -	struct page *page = pfn_to_page(pfn);
> -	struct folio *folio = page_folio(page);
> +	struct folio *folio = pfn_folio(pfn);
>   	LIST_HEAD(to_kill);
>   	dax_entry_t cookie;
>   	int rc = 0;
> 
> so I assume this is the v1->v3 delta.
> 

Yes.

> I'll queue this as a fixup patch with the changelog
> 
> add comment to mf_generic_kill_procss(), simplify
> mf_generic_kill_procs:folio initialization.
> 
> 
Thank you!
-jane

Patch

diff --git a/fs/dax.c b/fs/dax.c
index 906ecbd541a3..c70d4da047db 100644
--- a/fs/dax.c
+++ b/fs/dax.c
@@ -426,23 +426,23 @@  static struct page *dax_busy_page(void *entry)
 	return NULL;
 }
 
-/*
- * dax_lock_page - Lock the DAX entry corresponding to a page
- * @page: The page whose entry we want to lock
+/**
+ * dax_lock_folio - Lock the DAX entry corresponding to a folio
+ * @folio: The folio whose entry we want to lock
  *
  * Context: Process context.
- * Return: A cookie to pass to dax_unlock_page() or 0 if the entry could
+ * Return: A cookie to pass to dax_unlock_folio() or 0 if the entry could
  * not be locked.
  */
-dax_entry_t dax_lock_page(struct page *page)
+dax_entry_t dax_lock_folio(struct folio *folio)
 {
 	XA_STATE(xas, NULL, 0);
 	void *entry;
 
-	/* Ensure page->mapping isn't freed while we look at it */
+	/* Ensure folio->mapping isn't freed while we look at it */
 	rcu_read_lock();
 	for (;;) {
-		struct address_space *mapping = READ_ONCE(page->mapping);
+		struct address_space *mapping = READ_ONCE(folio->mapping);
 
 		entry = NULL;
 		if (!mapping || !dax_mapping(mapping))
@@ -461,11 +461,11 @@  dax_entry_t dax_lock_page(struct page *page)
 
 		xas.xa = &mapping->i_pages;
 		xas_lock_irq(&xas);
-		if (mapping != page->mapping) {
+		if (mapping != folio->mapping) {
 			xas_unlock_irq(&xas);
 			continue;
 		}
-		xas_set(&xas, page->index);
+		xas_set(&xas, folio->index);
 		entry = xas_load(&xas);
 		if (dax_is_locked(entry)) {
 			rcu_read_unlock();
@@ -481,10 +481,10 @@  dax_entry_t dax_lock_page(struct page *page)
 	return (dax_entry_t)entry;
 }
 
-void dax_unlock_page(struct page *page, dax_entry_t cookie)
+void dax_unlock_folio(struct folio *folio, dax_entry_t cookie)
 {
-	struct address_space *mapping = page->mapping;
-	XA_STATE(xas, &mapping->i_pages, page->index);
+	struct address_space *mapping = folio->mapping;
+	XA_STATE(xas, &mapping->i_pages, folio->index);
 
 	if (S_ISCHR(mapping->host->i_mode))
 		return;
diff --git a/include/linux/dax.h b/include/linux/dax.h
index 261944ec0887..711deb72c109 100644
--- a/include/linux/dax.h
+++ b/include/linux/dax.h
@@ -159,8 +159,8 @@  int dax_writeback_mapping_range(struct address_space *mapping,
 
 struct page *dax_layout_busy_page(struct address_space *mapping);
 struct page *dax_layout_busy_page_range(struct address_space *mapping, loff_t start, loff_t end);
-dax_entry_t dax_lock_page(struct page *page);
-void dax_unlock_page(struct page *page, dax_entry_t cookie);
+dax_entry_t dax_lock_folio(struct folio *folio);
+void dax_unlock_folio(struct folio *folio, dax_entry_t cookie);
 dax_entry_t dax_lock_mapping_entry(struct address_space *mapping,
 		unsigned long index, struct page **page);
 void dax_unlock_mapping_entry(struct address_space *mapping,
@@ -182,14 +182,14 @@  static inline int dax_writeback_mapping_range(struct address_space *mapping,
 	return -EOPNOTSUPP;
 }
 
-static inline dax_entry_t dax_lock_page(struct page *page)
+static inline dax_entry_t dax_lock_folio(struct folio *folio)
 {
-	if (IS_DAX(page->mapping->host))
+	if (IS_DAX(folio->mapping->host))
 		return ~0UL;
 	return 0;
 }
 
-static inline void dax_unlock_page(struct page *page, dax_entry_t cookie)
+static inline void dax_unlock_folio(struct folio *folio, dax_entry_t cookie)
 {
 }
 
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index fe121fdb05f7..6c2c9af0caa0 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -1710,20 +1710,23 @@  static void unmap_and_kill(struct list_head *to_kill, unsigned long pfn,
 	kill_procs(to_kill, flags & MF_MUST_KILL, false, pfn, flags);
 }
 
+/*
+ * Only dev_pagemap pages get here, such as fsdax when the filesystem
+ * either do not claim or fails to claim a hwpoison event, or devdax.
+ * The fsdax pages are initialized per base page, and the devdax pages
+ * could be initialized either as base pages, or as compound pages with
+ * vmemmap optimization enabled. Devdax is simplistic in its dealing with
+ * hwpoison, such that, if a subpage of a compound page is poisoned,
+ * simply mark the compound head page is by far sufficient.
+ */
 static int mf_generic_kill_procs(unsigned long long pfn, int flags,
 		struct dev_pagemap *pgmap)
 {
-	struct page *page = pfn_to_page(pfn);
+	struct folio *folio = pfn_folio(pfn);
 	LIST_HEAD(to_kill);
 	dax_entry_t cookie;
 	int rc = 0;
 
-	/*
-	 * Pages instantiated by device-dax (not filesystem-dax)
-	 * may be compound pages.
-	 */
-	page = compound_head(page);
-
 	/*
 	 * Prevent the inode from being freed while we are interrogating
 	 * the address_space, typically this would be handled by
@@ -1731,11 +1734,11 @@  static int mf_generic_kill_procs(unsigned long long pfn, int flags,
 	 * also prevents changes to the mapping of this pfn until
 	 * poison signaling is complete.
 	 */
-	cookie = dax_lock_page(page);
+	cookie = dax_lock_folio(folio);
 	if (!cookie)
 		return -EBUSY;
 
-	if (hwpoison_filter(page)) {
+	if (hwpoison_filter(&folio->page)) {
 		rc = -EOPNOTSUPP;
 		goto unlock;
 	}
@@ -1757,7 +1760,7 @@  static int mf_generic_kill_procs(unsigned long long pfn, int flags,
 	 * Use this flag as an indication that the dax page has been
 	 * remapped UC to prevent speculative consumption of poison.
 	 */
-	SetPageHWPoison(page);
+	SetPageHWPoison(&folio->page);
 
 	/*
 	 * Unlike System-RAM there is no possibility to swap in a
@@ -1766,11 +1769,11 @@  static int mf_generic_kill_procs(unsigned long long pfn, int flags,
 	 * SIGBUS (i.e. MF_MUST_KILL)
 	 */
 	flags |= MF_ACTION_REQUIRED | MF_MUST_KILL;
-	collect_procs(page, &to_kill, true);
+	collect_procs(&folio->page, &to_kill, true);
 
-	unmap_and_kill(&to_kill, pfn, page->mapping, page->index, flags);
+	unmap_and_kill(&to_kill, pfn, folio->mapping, folio->index, flags);
 unlock:
-	dax_unlock_page(page, cookie);
+	dax_unlock_folio(folio, cookie);
 	return rc;
 }