Received-SPF: pass (google.com: domain of linux-ext4+bounces-1429-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
From: Zhihao Cheng <chengzhihao1@huawei.com>
To: <brauner@kernel.org>, <david@fromorbit.com>, <djwong@kernel.org>,
	<jack@suse.cz>, <tytso@mit.edu>
CC: <linux-xfs@vger.kernel.org>, <linux-fsdevel@vger.kernel.org>,
	<linux-kernel@vger.kernel.org>, <linux-ext4@vger.kernel.org>,
	<yi.zhang@huawei.com>
Subject: [PATCH RFC 1/2] iomap: Add a IOMAP_DIO_MAY_INLINE_COMP flag
Date: Thu, 29 Feb 2024 19:38:48 +0800
Message-ID: <20240229113849.2222577-2-chengzhihao1@huawei.com>
In-Reply-To: <20240229113849.2222577-1-chengzhihao1@huawei.com>
References: <20240229113849.2222577-1-chengzhihao1@huawei.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Content-Type: text/plain

It will be more efficient to execute quick endio process(eg. non-sync
overwriting case) under irq process rather than starting a worker to
do it.
Add a flag to control DIO to be finished inline(under irq context), which
can be used for non-sync overwriting case.
Besides, skip invalidating pages if DIO is finished inline, which will
keep the same logic with dio_bio_end_aio in non-sync overwriting case.

Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
---
 fs/iomap/direct-io.c  | 10 ++++++++--
 include/linux/iomap.h |  6 ++++++
 2 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
index bcd3f8cf5ea4..221715b38ce2 100644
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -110,7 +110,8 @@ ssize_t iomap_dio_complete(struct iomap_dio *dio)
 	 * ->end_io() when necessary, otherwise a racing buffer read would cache
 	 * zeros from unwritten extents.
 	 */
-	if (!dio->error && dio->size && (dio->flags & IOMAP_DIO_WRITE))
+	if (!dio->error && dio->size && (dio->flags & IOMAP_DIO_WRITE) &&
+	    !(dio->flags & IOMAP_DIO_INLINE_COMP))
 		kiocb_invalidate_post_direct_write(iocb, dio->size);
 
 	inode_dio_end(file_inode(iocb->ki_filp));
@@ -122,8 +123,10 @@ ssize_t iomap_dio_complete(struct iomap_dio *dio)
 		 * If this is a DSYNC write, make sure we push it to stable
 		 * storage now that we've written data.
 		 */
-		if (dio->flags & IOMAP_DIO_NEED_SYNC)
+		if (dio->flags & IOMAP_DIO_NEED_SYNC) {
+			WARN_ON_ONCE(dio->flags & IOMAP_DIO_INLINE_COMP);
 			ret = generic_write_sync(iocb, ret);
+		}
 		if (ret > 0)
 			ret += dio->done_before;
 	}
@@ -628,6 +631,9 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 			*/
 			if (!(iocb->ki_flags & IOCB_SYNC))
 				dio->flags |= IOMAP_DIO_WRITE_THROUGH;
+		} else if (dio_flags & IOMAP_DIO_MAY_INLINE_COMP) {
+			/* writes could complete inline */
+			dio->flags |= IOMAP_DIO_INLINE_COMP;
 		}
 
 		/*
diff --git a/include/linux/iomap.h b/include/linux/iomap.h
index 96dd0acbba44..f292b10028d0 100644
--- a/include/linux/iomap.h
+++ b/include/linux/iomap.h
@@ -382,6 +382,12 @@ struct iomap_dio_ops {
  */
 #define IOMAP_DIO_PARTIAL		(1 << 2)
 
+/*
+ * DIO will be completed inline unless sync operation is needed after io is
+ * finished.
+ */
+#define IOMAP_DIO_MAY_INLINE_COMP	(1 << 3)
+
 ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 		const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
 		unsigned int dio_flags, void *private, size_t done_before);
-- 
2.39.2