diff mbox series

[v6,02/30] mm: Add set/end/wait functions for PG_private_2

Message ID 161789066013.6155.9816857201817288382.stgit@warthog.procyon.org.uk (mailing list archive)
State New
Headers show
Series Network fs helper library & fscache kiocb API | expand

Commit Message

David Howells April 8, 2021, 2:04 p.m. UTC
Add three functions to manipulate PG_private_2:

 (*) set_page_private_2() - Set the flag and take an appropriate reference
     on the flagged page.

 (*) end_page_private_2() - Clear the flag, drop the reference and wake up
     any waiters, somewhat analogously with end_page_writeback().

 (*) wait_on_page_private_2() - Wait for the flag to be cleared.

Wrappers will need to be placed in the netfs lib header in the patch that
adds that.

[This implements a suggestion by Linus[1] to not mix the terminology of
 PG_private_2 and PG_fscache in the mm core function]

Changes:
v5:
- Add set and end functions, calling the end function end rather than
  unlock[3].
- Keep a ref on the page when PG_private_2 is set[4][5].

v4:
- Remove extern from the declaration[2].

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Matthew Wilcox (Oracle) <willy@infradead.org>
cc: Alexander Viro <viro@zeniv.linux.org.uk>
cc: Christoph Hellwig <hch@lst.de>
cc: linux-mm@kvack.org
cc: linux-cachefs@redhat.com
cc: linux-afs@lists.infradead.org
cc: linux-nfs@vger.kernel.org
cc: linux-cifs@vger.kernel.org
cc: ceph-devel@vger.kernel.org
cc: v9fs-developer@lists.sourceforge.net
cc: linux-fsdevel@vger.kernel.org
Link: https://lore.kernel.org/r/1330473.1612974547@warthog.procyon.org.uk/ # v1
Link: https://lore.kernel.org/r/CAHk-=wjgA-74ddehziVk=XAEMTKswPu1Yw4uaro1R3ibs27ztw@mail.gmail.com/ [1]
Link: https://lore.kernel.org/r/20210216102659.GA27714@lst.de/ [2]
Link: https://lore.kernel.org/r/161340387944.1303470.7944159520278177652.stgit@warthog.procyon.org.uk/ # v3
Link: https://lore.kernel.org/r/161539528910.286939.1252328699383291173.stgit@warthog.procyon.org.uk # v4
Link: https://lore.kernel.org/r/20210321105309.GG3420@casper.infradead.org [3]
Link: https://lore.kernel.org/r/CAHk-=wh+2gbF7XEjYc=HV9w_2uVzVf7vs60BPz0gFA=+pUm3ww@mail.gmail.com/ [4]
Link: https://lore.kernel.org/r/CAHk-=wjSGsRj7xwhSMQ6dAQiz53xA39pOG+XA_WeTgwBBu4uqg@mail.gmail.com/ [5]
Link: https://lore.kernel.org/r/161653788200.2770958.9517755716374927208.stgit@warthog.procyon.org.uk/ # v5
---

 include/linux/pagemap.h |   19 +++++++++++++++
 mm/filemap.c            |   59 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 78 insertions(+)

Comments

Matthew Wilcox April 8, 2021, 2:50 p.m. UTC | #1
On Thu, Apr 08, 2021 at 03:04:20PM +0100, David Howells wrote:
> +static inline void set_page_private_2(struct page *page)
> +{
> +	get_page(page);
> +	SetPagePrivate2(page);

PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)

So we can set Private2 on any tail page ...

> +void end_page_private_2(struct page *page)
> +{
> +	page = compound_head(page);
> +	VM_BUG_ON_PAGE(!PagePrivate2(page), page);
> +	clear_bit_unlock(PG_private_2, &page->flags);
> +	wake_up_page_bit(page, PG_private_2);

... but when we try to end on a tail, we actually wake up the head ...

> +void wait_on_page_private_2(struct page *page)
> +{
> +	while (PagePrivate2(page))
> +		wait_on_page_bit(page, PG_private_2);

... although if we were waiting on a tail, the wake up won't find us ...

if only we had a way to ensure this kind of bug can't happen *cough,
lend your support to the page folio patches*.
David Howells April 8, 2021, 3:25 p.m. UTC | #2
Matthew Wilcox <willy@infradead.org> wrote:

> > +void end_page_private_2(struct page *page)
> > +{
> > +	page = compound_head(page);
> > +	VM_BUG_ON_PAGE(!PagePrivate2(page), page);
> > +	clear_bit_unlock(PG_private_2, &page->flags);
> > +	wake_up_page_bit(page, PG_private_2);
> 
> ... but when we try to end on a tail, we actually wake up the head ...

Question is, should I remove compound_head() here or add it into the other
functions?

David
David Howells April 8, 2021, 3:57 p.m. UTC | #3
Here's a partial change, but we still need to deal with the assumption that
page_has_private() makes that its output can be used to count the number of
refs held for PG_private *and* PG_private_2 - which isn't true for my code
here.

David
---
commit e7c28d83b84b972c3faa0dd86020548aa50eda75
Author: David Howells <dhowells@redhat.com>
Date:   Thu Apr 8 16:33:20 2021 +0100

    netfs: Fix PG_private_2 helper functions to consistently use compound_head()

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index ef511364cc0c..63ca6430aef5 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -699,6 +699,7 @@ void page_endio(struct page *page, bool is_write, int err);
  */
 static inline void set_page_private_2(struct page *page)
 {
+	page = compound_head(page);
 	get_page(page);
 	SetPagePrivate2(page);
 }
diff --git a/mm/filemap.c b/mm/filemap.c
index 0ce93c8799ca..46e0321ba87a 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1461,6 +1461,7 @@ EXPORT_SYMBOL(end_page_private_2);
  */
 void wait_on_page_private_2(struct page *page)
 {
+	page = compound_head(page);
 	while (PagePrivate2(page))
 		wait_on_page_bit(page, PG_private_2);
 }
@@ -1481,6 +1482,7 @@ int wait_on_page_private_2_killable(struct page *page)
 {
 	int ret = 0;
 
+	page = compound_head(page);
 	while (PagePrivate2(page)) {
 		ret = wait_on_page_bit_killable(page, PG_private_2);
 		if (ret < 0)
diff mbox series

Patch

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 8c9947fd62f3..4a7c916abb5c 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -688,6 +688,25 @@  void wait_for_stable_page(struct page *page);
 
 void page_endio(struct page *page, bool is_write, int err);
 
+/**
+ * set_page_private_2 - Set PG_private_2 on a page and take a ref
+ * @page: The page.
+ *
+ * Set the PG_private_2 flag on a page and take the reference needed for the VM
+ * to handle its lifetime correctly.  This sets the flag and takes the
+ * reference unconditionally, so care must be taken not to set the flag again
+ * if it's already set.
+ */
+static inline void set_page_private_2(struct page *page)
+{
+	get_page(page);
+	SetPagePrivate2(page);
+}
+
+void end_page_private_2(struct page *page);
+void wait_on_page_private_2(struct page *page);
+int wait_on_page_private_2_killable(struct page *page);
+
 /*
  * Add an arbitrary waiter to a page's wait queue
  */
diff --git a/mm/filemap.c b/mm/filemap.c
index 43700480d897..788b71e8a72d 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1432,6 +1432,65 @@  void unlock_page(struct page *page)
 }
 EXPORT_SYMBOL(unlock_page);
 
+/**
+ * end_page_private_2 - Clear PG_private_2 and release any waiters
+ * @page: The page
+ *
+ * Clear the PG_private_2 bit on a page and wake up any sleepers waiting for
+ * this.  The page ref held for PG_private_2 being set is released.
+ *
+ * This is, for example, used when a netfs page is being written to a local
+ * disk cache, thereby allowing writes to the cache for the same page to be
+ * serialised.
+ */
+void end_page_private_2(struct page *page)
+{
+	page = compound_head(page);
+	VM_BUG_ON_PAGE(!PagePrivate2(page), page);
+	clear_bit_unlock(PG_private_2, &page->flags);
+	wake_up_page_bit(page, PG_private_2);
+	put_page(page);
+}
+EXPORT_SYMBOL(end_page_private_2);
+
+/**
+ * wait_on_page_private_2 - Wait for PG_private_2 to be cleared on a page
+ * @page: The page to wait on
+ *
+ * Wait for PG_private_2 (aka PG_fscache) to be cleared on a page.
+ */
+void wait_on_page_private_2(struct page *page)
+{
+	while (PagePrivate2(page))
+		wait_on_page_bit(page, PG_private_2);
+}
+EXPORT_SYMBOL(wait_on_page_private_2);
+
+/**
+ * wait_on_page_private_2_killable - Wait for PG_private_2 to be cleared on a page
+ * @page: The page to wait on
+ *
+ * Wait for PG_private_2 (aka PG_fscache) to be cleared on a page or until a
+ * fatal signal is received by the calling task.
+ *
+ * Return:
+ * - 0 if successful.
+ * - -EINTR if a fatal signal was encountered.
+ */
+int wait_on_page_private_2_killable(struct page *page)
+{
+	int ret = 0;
+
+	while (PagePrivate2(page)) {
+		ret = wait_on_page_bit_killable(page, PG_private_2);
+		if (ret < 0)
+			break;
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(wait_on_page_private_2_killable);
+
 /**
  * end_page_writeback - end writeback against a page
  * @page: the page