diff mbox

[03/11] fs: add support for an inode to carry stream related data

Message ID 1497498312-17704-4-git-send-email-axboe@kernel.dk (mailing list archive)
State New, archived
Headers show

Commit Message

Jens Axboe June 15, 2017, 3:45 a.m. UTC 
No functional changes in this patch, just in preparation for
allowing applications to pass in hints about data life times
for writes.

Pack the i_write_hint field into a 2-byte hole, so we don't grow
the size of the inode.

Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/inode.c         | 1 +
 include/linux/fs.h | 9 +++++++++
 2 files changed, 10 insertions(+)

Comments

Christoph Hellwig June 15, 2017, 8:17 a.m. UTC | #1 
On Wed, Jun 14, 2017 at 09:45:04PM -0600, Jens Axboe wrote:
> No functional changes in this patch, just in preparation for
> allowing applications to pass in hints about data life times
> for writes.
> 
> Pack the i_write_hint field into a 2-byte hole, so we don't grow
> the size of the inode.

A u8 should be plenty.  But talking about the representation -
your write lifetime hints are a 5 option enum basically.  I wonder
if we really should encode it as flags, or if we should have an
enum (which could be packed into a 3-bit bitfield) and then pass
it down the stack in that form instead of changing the representation
N times.
Jens Axboe June 15, 2017, 2:22 p.m. UTC | #2 
On 06/15/2017 02:17 AM, Christoph Hellwig wrote:
> On Wed, Jun 14, 2017 at 09:45:04PM -0600, Jens Axboe wrote:
>> No functional changes in this patch, just in preparation for
>> allowing applications to pass in hints about data life times
>> for writes.
>>
>> Pack the i_write_hint field into a 2-byte hole, so we don't grow
>> the size of the inode.
> 
> A u8 should be plenty.  But talking about the representation -
> your write lifetime hints are a 5 option enum basically.  I wonder
> if we really should encode it as flags, or if we should have an
> enum (which could be packed into a 3-bit bitfield) and then pass
> it down the stack in that form instead of changing the representation
> N times.

If we keep the RWF_WRITE_LIFE_* flags, then yes, I think we should
unify the RWF_WRITE_LIFE_*, IOCB_WRITE_LIFE_*, and REQ_WRITE_LIFE_*
flags into a specific type.
diff mbox

Patch

diff --git a/fs/inode.c b/fs/inode.c
index db5914783a71..bd8bf44f3f31 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -149,6 +149,7 @@  int inode_init_always(struct super_block *sb, struct inode *inode)
 	inode->i_blocks = 0;
 	inode->i_bytes = 0;
 	inode->i_generation = 0;
+	inode->i_write_hint = 0;
 	inode->i_pipe = NULL;
 	inode->i_bdev = NULL;
 	inode->i_cdev = NULL;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 803e5a9b2654..f4f9df8ed059 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -591,6 +591,7 @@  struct inode {
 	struct timespec		i_ctime;
 	spinlock_t		i_lock;	/* i_blocks, i_bytes, maybe i_size */
 	unsigned short          i_bytes;
+	unsigned short		i_write_hint;
 	unsigned int		i_blkbits;
 	blkcnt_t		i_blocks;
 
@@ -655,6 +656,14 @@  struct inode {
 	void			*i_private; /* fs or device private pointer */
 };
 
+static inline unsigned int inode_write_hint(struct inode *inode)
+{
+	if (inode)
+		return inode->i_write_hint;
+
+	return 0;
+}
+
 static inline unsigned int i_blocksize(const struct inode *node)
 {
 	return (1 << node->i_blkbits);