[PATCH v4 0/3] Initial NFS client support for RWF_DONTCACHE

[PATCH v4 0/3] Initial NFS client support for RWF_DONTCACHE

[Trond Myklebust]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

The following patch set attempts to add support for the RWF_DONTCACHE
flag in preadv2() and pwritev2() on NFS filesystems.

The main issue is allowing support on 2 stage writes (i.e. unstable
WRITE followed by a COMMIT) since those don't follow the current
assumption that the 'dropbehind' flag can be fulfilled as soon as the
writeback lock is dropped.

v2:
 - Make use of the new iocb parameter for nfs_write_begin()
v3:
 - Set/clear PG_DROPBEHIND on the head of the nfs_page group
 - Simplify helper folio_end_dropbehind
v4:
 - Replace filemap_end_dropbehind_write() with folio_end_dropbehind()
 - Add a helper to replace folio_end_writeback with an equivalent that
   does not attempt to interpret the dropbehind flag
 - Keep the folio dropbehind flag set until the NFS client is ready to
   call folio_end_dropbehind.
 - Don't try to do a read-modify-write in nfs_write_begin() if the folio
   has the dropbehind flag set.

Trond Myklebust (3):
  filemap: Add a helper for filesystems implementing dropbehind
  filemap: Add a version of folio_end_writeback that ignores dropbehind
  NFS: Enable use of the RWF_DONTCACHE flag on the NFS client

 fs/nfs/file.c           |  9 +++++----
 fs/nfs/nfs4file.c       |  1 +
 fs/nfs/write.c          |  4 +++-
 include/linux/pagemap.h |  2 ++
 mm/filemap.c            | 34 ++++++++++++++++++++++++++--------
 5 files changed, 37 insertions(+), 13 deletions(-)

-- 
2.50.1

[PATCH v4 1/3] filemap: Add a helper for filesystems implementing dropbehind

[Trond Myklebust]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

Add a helper to allow filesystems to attempt to free the 'dropbehind'
folio.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Link: https://lore.kernel.org/all/5588a06f6d5a2cf6746828e2d36e7ada668b1739.1745381692.git.trond.myklebust@hammerspace.com/
Reviewed-by: Mike Snitzer <snitzer@kernel.org>
---
 include/linux/pagemap.h | 1 +
 mm/filemap.c            | 5 +++--
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 12a12dae727d..201b7c6f6441 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -1221,6 +1221,7 @@ void folio_wait_writeback(struct folio *folio);
 int folio_wait_writeback_killable(struct folio *folio);
 void end_page_writeback(struct page *page);
 void folio_end_writeback(struct folio *folio);
+void folio_end_dropbehind(struct folio *folio);
 void folio_wait_stable(struct folio *folio);
 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
diff --git a/mm/filemap.c b/mm/filemap.c
index 751838ef05e5..71209ebbcc36 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1608,7 +1608,7 @@ static void filemap_end_dropbehind(struct folio *folio)
  * completes. Do that now. If we fail, it's likely because of a big folio -
  * just reset dropbehind for that case and latter completions should invalidate.
  */
-static void filemap_end_dropbehind_write(struct folio *folio)
+void folio_end_dropbehind(struct folio *folio)
 {
 	if (!folio_test_dropbehind(folio))
 		return;
@@ -1625,6 +1625,7 @@ static void filemap_end_dropbehind_write(struct folio *folio)
 		folio_unlock(folio);
 	}
 }
+EXPORT_SYMBOL(folio_end_dropbehind);
 
 /**
  * folio_end_writeback - End writeback against a folio.
@@ -1660,7 +1661,7 @@ void folio_end_writeback(struct folio *folio)
 	if (__folio_end_writeback(folio))
 		folio_wake_bit(folio, PG_writeback);
 
-	filemap_end_dropbehind_write(folio);
+	folio_end_dropbehind(folio);
 	acct_reclaim_writeback(folio);
 	folio_put(folio);
 }
-- 
2.50.1

[PATCH v4 2/3] filemap: Add a version of folio_end_writeback that ignores dropbehind

[Trond Myklebust]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

Filesystems such as NFS may need to defer dropbehind until after their
2-stage writes are done. This adds a helper
folio_end_writeback_no_dropbehind() that allows them to release the
writeback flag without immediately dropping the folio.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 include/linux/pagemap.h |  1 +
 mm/filemap.c            | 29 +++++++++++++++++++++++------
 2 files changed, 24 insertions(+), 6 deletions(-)

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 201b7c6f6441..5b26465358ce 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -1221,6 +1221,7 @@ void folio_wait_writeback(struct folio *folio);
 int folio_wait_writeback_killable(struct folio *folio);
 void end_page_writeback(struct page *page);
 void folio_end_writeback(struct folio *folio);
+void folio_end_writeback_no_dropbehind(struct folio *folio);
 void folio_end_dropbehind(struct folio *folio);
 void folio_wait_stable(struct folio *folio);
 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
diff --git a/mm/filemap.c b/mm/filemap.c
index 71209ebbcc36..0cf147ca7c16 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1628,14 +1628,15 @@ void folio_end_dropbehind(struct folio *folio)
 EXPORT_SYMBOL(folio_end_dropbehind);
 
 /**
- * folio_end_writeback - End writeback against a folio.
+ * folio_end_writeback_no_dropbehind - End writeback against a folio.
  * @folio: The folio.
  *
  * The folio must actually be under writeback.
+ * This call is intended for filesystems that need to defer dropbehind.
  *
  * Context: May be called from process or interrupt context.
  */
-void folio_end_writeback(struct folio *folio)
+void folio_end_writeback_no_dropbehind(struct folio *folio)
 {
 	VM_BUG_ON_FOLIO(!folio_test_writeback(folio), folio);
 
@@ -1651,6 +1652,25 @@ void folio_end_writeback(struct folio *folio)
 		folio_rotate_reclaimable(folio);
 	}
 
+	if (__folio_end_writeback(folio))
+		folio_wake_bit(folio, PG_writeback);
+
+	acct_reclaim_writeback(folio);
+}
+EXPORT_SYMBOL(folio_end_writeback_no_dropbehind);
+
+/**
+ * folio_end_writeback - End writeback against a folio.
+ * @folio: The folio.
+ *
+ * The folio must actually be under writeback.
+ *
+ * Context: May be called from process or interrupt context.
+ */
+void folio_end_writeback(struct folio *folio)
+{
+	VM_BUG_ON_FOLIO(!folio_test_writeback(folio), folio);
+
 	/*
 	 * Writeback does not hold a folio reference of its own, relying
 	 * on truncation to wait for the clearing of PG_writeback.
@@ -1658,11 +1678,8 @@ void folio_end_writeback(struct folio *folio)
 	 * reused before the folio_wake_bit().
 	 */
 	folio_get(folio);
-	if (__folio_end_writeback(folio))
-		folio_wake_bit(folio, PG_writeback);
-
+	folio_end_writeback_no_dropbehind(folio);
 	folio_end_dropbehind(folio);
-	acct_reclaim_writeback(folio);
 	folio_put(folio);
 }
 EXPORT_SYMBOL(folio_end_writeback);
-- 
2.50.1

[PATCH v4 3/3] NFS: Enable use of the RWF_DONTCACHE flag on the NFS client

[Trond Myklebust]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

The NFS client needs to defer dropbehind until after any writes to the
folio have been persisted on the server. Since this may be a 2 step
process, use folio_end_writeback_no_dropbehind() to allow release of the
writeback flag, and then call folio_end_dropbehind() once the COMMIT is
done.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/file.c     | 9 +++++----
 fs/nfs/nfs4file.c | 1 +
 fs/nfs/write.c    | 4 +++-
 3 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 86e36c630f09..90610629862a 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -329,6 +329,8 @@ static bool nfs_want_read_modify_write(struct file *file, struct folio *folio,
 
 	if (pnfs_ld_read_whole_page(file_inode(file)))
 		return true;
+	if (folio_test_dropbehind(folio))
+		return false;
 	/* Open for reading too? */
 	if (file->f_mode & FMODE_READ)
 		return true;
@@ -348,7 +350,6 @@ static int nfs_write_begin(const struct kiocb *iocb,
 			   loff_t pos, unsigned len, struct folio **foliop,
 			   void **fsdata)
 {
-	fgf_t fgp = FGP_WRITEBEGIN;
 	struct folio *folio;
 	struct file *file = iocb->ki_filp;
 	int once_thru = 0;
@@ -357,10 +358,8 @@ static int nfs_write_begin(const struct kiocb *iocb,
 	dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n",
 		file, mapping->host->i_ino, len, (long long) pos);
 
-	fgp |= fgf_set_order(len);
 start:
-	folio = __filemap_get_folio(mapping, pos >> PAGE_SHIFT, fgp,
-				    mapping_gfp_mask(mapping));
+	folio = write_begin_get_folio(iocb, mapping, pos >> PAGE_SHIFT, len);
 	if (IS_ERR(folio))
 		return PTR_ERR(folio);
 	*foliop = folio;
@@ -372,6 +371,7 @@ static int nfs_write_begin(const struct kiocb *iocb,
 	} else if (!once_thru &&
 		   nfs_want_read_modify_write(file, folio, pos, len)) {
 		once_thru = 1;
+		folio_clear_dropbehind(folio);
 		ret = nfs_read_folio(file, folio);
 		folio_put(folio);
 		if (!ret)
@@ -915,5 +915,6 @@ const struct file_operations nfs_file_operations = {
 	.splice_write	= iter_file_splice_write,
 	.check_flags	= nfs_check_flags,
 	.setlease	= simple_nosetlease,
+	.fop_flags	= FOP_DONTCACHE,
 };
 EXPORT_SYMBOL_GPL(nfs_file_operations);
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index 1d6b5f4230c9..654fda6f362c 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -454,4 +454,5 @@ const struct file_operations nfs4_file_operations = {
 #else
 	.llseek		= nfs_file_llseek,
 #endif
+	.fop_flags	= FOP_DONTCACHE,
 };
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 8b7c04737967..079fc1af928f 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -338,7 +338,7 @@ static void nfs_folio_end_writeback(struct folio *folio)
 {
 	struct nfs_server *nfss = NFS_SERVER(folio->mapping->host);
 
-	folio_end_writeback(folio);
+	folio_end_writeback_no_dropbehind(folio);
 	if (atomic_long_dec_return(&nfss->writeback) <
 	    NFS_CONGESTION_OFF_THRESH) {
 		nfss->write_congested = 0;
@@ -787,6 +787,8 @@ static void nfs_inode_remove_request(struct nfs_page *req)
 			clear_bit(PG_MAPPED, &req->wb_head->wb_flags);
 		}
 		spin_unlock(&mapping->i_private_lock);
+
+		folio_end_dropbehind(folio);
 	}
 	nfs_page_group_unlock(req);
 
-- 
2.50.1

Re: [PATCH v4 1/3] filemap: Add a helper for filesystems implementing dropbehind

[Christoph Hellwig]

On Tue, Aug 19, 2025 at 07:18:30AM -0700, Trond Myklebust wrote:
> +EXPORT_SYMBOL(folio_end_dropbehind);

This really should be a EXPORT_SYMBOL_GPL like for any other linux-only
functionality. Same for the next patch.