[PATCH v6 00/16] xfs: Delay Ready Attributes

[PATCH v6 00/16] xfs: Delay Ready Attributes

[Allison Collins]

Hi all,

This set is a subset of a larger series for delayed attributes. Which is a
subset of an even larger series, parent pointers. Delayed attributes allow
attribute operations (set and remove) to be logged and committed in the same way
that other delayed operations do. This allows more complex operations (like
parent pointers) to be broken up into multiple smaller transactions. To do this,
the existing attr operations must be modified to operate as either a delayed
operation or a inline operation since older filesystems will not be able to use
the new log entries. This means that they cannot roll, commit, or finish
transactions. Instead, they return  EAGAIN an allow the calling function to
handle the transaction. In this series, we focus on only the clean up and
refactoring needed to accomplish this. We will introduce delayed attrs and
parent pointers in a later set.

I know there are some other sets waiting for review that intersect with this
one, and I still need to go through those (not forgotten, just queued).  But I
wanted to get an update out before too much time gets away.

Changes since v5:
Mostly review updates and nits collected since v5.  Patches 12, 13 and 14 are
new, and help to modularize and simplify the larger routines in 15 and 16.

I've also made the corresponding updates to the user space side as well.

Questions, comment and feedback appreciated! 

Thanks all!
Allison

Allison Collins (15):
  xfs: Replace attribute parameters with struct xfs_name
  xfs: Embed struct xfs_name in xfs_da_args
  xfs: Factor out new helper functions xfs_attr_rmtval_set
  xfs: Factor out trans handling in xfs_attr3_leaf_flipflags
  xfs: Factor out xfs_attr_leaf_addname helper
  xfs: Refactor xfs_attr_try_sf_addname
  xfs: Factor out trans roll from xfs_attr3_leaf_setflag
  xfs: Factor out xfs_attr_rmtval_invalidate
  xfs: Factor out trans roll in xfs_attr3_leaf_clearflag
  xfs: Check for -ENOATTR or -EEXIST
  xfs: Add helper function xfs_attr_init_unmapstate
  xfs: Add helper function xfs_attr_rmtval_unmap
  xfs: Simplify xfs_attr_set_args
  xfs: Add delay ready attr remove routines
  xfs: Add delay ready attr set routines

Allison Henderson (1):
  xfs: Add xfs_has_attr and subroutines

 fs/xfs/libxfs/xfs_attr.c        | 866 +++++++++++++++++++++++++++-------------
 fs/xfs/libxfs/xfs_attr.h        |  15 +-
 fs/xfs/libxfs/xfs_attr_leaf.c   | 220 +++++-----
 fs/xfs/libxfs/xfs_attr_leaf.h   |   3 +
 fs/xfs/libxfs/xfs_attr_remote.c | 265 +++++++++---
 fs/xfs/libxfs/xfs_attr_remote.h |   7 +-
 fs/xfs/libxfs/xfs_da_btree.c    |   6 +-
 fs/xfs/libxfs/xfs_da_btree.h    |  41 +-
 fs/xfs/libxfs/xfs_dir2.c        |  18 +-
 fs/xfs/libxfs/xfs_dir2_block.c  |   6 +-
 fs/xfs/libxfs/xfs_dir2_leaf.c   |   6 +-
 fs/xfs/libxfs/xfs_dir2_node.c   |   8 +-
 fs/xfs/libxfs/xfs_dir2_sf.c     |  30 +-
 fs/xfs/libxfs/xfs_types.c       |  11 +
 fs/xfs/libxfs/xfs_types.h       |   1 +
 fs/xfs/scrub/attr.c             |  12 +-
 fs/xfs/scrub/common.c           |   2 +
 fs/xfs/xfs_acl.c                |  27 +-
 fs/xfs/xfs_attr_list.c          |   1 +
 fs/xfs/xfs_ioctl.c              |  25 +-
 fs/xfs/xfs_ioctl32.c            |   2 +
 fs/xfs/xfs_iops.c               |   8 +-
 fs/xfs/xfs_trace.h              |  20 +-
 fs/xfs/xfs_xattr.c              |  29 +-
 24 files changed, 1095 insertions(+), 534 deletions(-)

-- 
2.7.4

[PATCH v6 01/16] xfs: Replace attribute parameters with struct xfs_name

[Allison Collins]

This patch replaces the attribute name and length parameters with a
single struct xfs_name parameter.  This helps to clean up the numbers of
parameters being passed around and pre-simplifies the code some.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c  | 22 +++++++++-------------
 fs/xfs/libxfs/xfs_attr.h  | 12 +++++-------
 fs/xfs/libxfs/xfs_types.c | 11 +++++++++++
 fs/xfs/libxfs/xfs_types.h |  1 +
 fs/xfs/xfs_acl.c          | 25 +++++++++++--------------
 fs/xfs/xfs_ioctl.c        | 23 +++++++++++++----------
 fs/xfs/xfs_iops.c         |  6 +++---
 fs/xfs/xfs_xattr.c        | 28 ++++++++++++++++------------
 8 files changed, 69 insertions(+), 59 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index e614972..6717f47 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -61,8 +61,7 @@ STATIC int
 xfs_attr_args_init(
 	struct xfs_da_args	*args,
 	struct xfs_inode	*dp,
-	const unsigned char	*name,
-	size_t			namelen,
+	struct xfs_name		*name,
 	int			flags)
 {
 
@@ -74,8 +73,8 @@ xfs_attr_args_init(
 	args->whichfork = XFS_ATTR_FORK;
 	args->dp = dp;
 	args->flags = flags;
-	args->name = name;
-	args->namelen = namelen;
+	args->name = name->name;
+	args->namelen = name->len;
 	if (args->namelen >= MAXNAMELEN)
 		return -EFAULT;		/* match IRIX behaviour */
 
@@ -139,8 +138,7 @@ xfs_attr_get_ilocked(
 int
 xfs_attr_get(
 	struct xfs_inode	*ip,
-	const unsigned char	*name,
-	size_t			namelen,
+	struct xfs_name		*name,
 	unsigned char		**value,
 	int			*valuelenp,
 	int			flags)
@@ -156,7 +154,7 @@ xfs_attr_get(
 	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
 		return -EIO;
 
-	error = xfs_attr_args_init(&args, ip, name, namelen, flags);
+	error = xfs_attr_args_init(&args, ip, name, flags);
 	if (error)
 		return error;
 
@@ -339,8 +337,7 @@ xfs_attr_remove_args(
 int
 xfs_attr_set(
 	struct xfs_inode	*dp,
-	const unsigned char	*name,
-	size_t			namelen,
+	struct xfs_name		*name,
 	unsigned char		*value,
 	int			valuelen,
 	int			flags)
@@ -356,7 +353,7 @@ xfs_attr_set(
 	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
 		return -EIO;
 
-	error = xfs_attr_args_init(&args, dp, name, namelen, flags);
+	error = xfs_attr_args_init(&args, dp, name, flags);
 	if (error)
 		return error;
 
@@ -444,8 +441,7 @@ xfs_attr_set(
 int
 xfs_attr_remove(
 	struct xfs_inode	*dp,
-	const unsigned char	*name,
-	size_t			namelen,
+	struct xfs_name		*name,
 	int			flags)
 {
 	struct xfs_mount	*mp = dp->i_mount;
@@ -457,7 +453,7 @@ xfs_attr_remove(
 	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
 		return -EIO;
 
-	error = xfs_attr_args_init(&args, dp, name, namelen, flags);
+	error = xfs_attr_args_init(&args, dp, name, flags);
 	if (error)
 		return error;
 
diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
index 4243b22..35043db 100644
--- a/fs/xfs/libxfs/xfs_attr.h
+++ b/fs/xfs/libxfs/xfs_attr.h
@@ -147,14 +147,12 @@ int xfs_attr_list_int_ilocked(struct xfs_attr_list_context *);
 int xfs_attr_list_int(struct xfs_attr_list_context *);
 int xfs_inode_hasattr(struct xfs_inode *ip);
 int xfs_attr_get_ilocked(struct xfs_inode *ip, struct xfs_da_args *args);
-int xfs_attr_get(struct xfs_inode *ip, const unsigned char *name,
-		 size_t namelen, unsigned char **value, int *valuelenp,
-		 int flags);
-int xfs_attr_set(struct xfs_inode *dp, const unsigned char *name,
-		 size_t namelen, unsigned char *value, int valuelen, int flags);
+int xfs_attr_get(struct xfs_inode *ip, struct xfs_name *name,
+		 unsigned char **value, int *valuelenp, int flags);
+int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
+		 unsigned char *value, int valuelen, int flags);
 int xfs_attr_set_args(struct xfs_da_args *args);
-int xfs_attr_remove(struct xfs_inode *dp, const unsigned char *name,
-		    size_t namelen, int flags);
+int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
 int xfs_attr_remove_args(struct xfs_da_args *args);
 int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
 		  int flags, struct attrlist_cursor_kern *cursor);
diff --git a/fs/xfs/libxfs/xfs_types.c b/fs/xfs/libxfs/xfs_types.c
index 4f59554..781a5a9 100644
--- a/fs/xfs/libxfs/xfs_types.c
+++ b/fs/xfs/libxfs/xfs_types.c
@@ -12,6 +12,17 @@
 #include "xfs_bit.h"
 #include "xfs_mount.h"
 
+/* Initialize a struct xfs_name with a null terminated string name */
+void
+xfs_name_init(
+	struct xfs_name *xname,
+	const char *name)
+{
+	xname->name = (unsigned char *)name;
+	xname->len = strlen(name);
+	xname->type = 0;
+}
+
 /* Find the size of the AG, in blocks. */
 xfs_agblock_t
 xfs_ag_block_count(
diff --git a/fs/xfs/libxfs/xfs_types.h b/fs/xfs/libxfs/xfs_types.h
index 397d947..b94acb5 100644
--- a/fs/xfs/libxfs/xfs_types.h
+++ b/fs/xfs/libxfs/xfs_types.h
@@ -180,6 +180,7 @@ enum xfs_ag_resv_type {
  */
 struct xfs_mount;
 
+void xfs_name_init(struct xfs_name *xname, const char *name);
 xfs_agblock_t xfs_ag_block_count(struct xfs_mount *mp, xfs_agnumber_t agno);
 bool xfs_verify_agbno(struct xfs_mount *mp, xfs_agnumber_t agno,
 		xfs_agblock_t agbno);
diff --git a/fs/xfs/xfs_acl.c b/fs/xfs/xfs_acl.c
index cd743fad..42ac847 100644
--- a/fs/xfs/xfs_acl.c
+++ b/fs/xfs/xfs_acl.c
@@ -123,7 +123,7 @@ xfs_get_acl(struct inode *inode, int type)
 	struct xfs_inode *ip = XFS_I(inode);
 	struct posix_acl *acl = NULL;
 	struct xfs_acl *xfs_acl = NULL;
-	unsigned char *ea_name;
+	struct xfs_name name;
 	int error;
 	int len;
 
@@ -131,10 +131,10 @@ xfs_get_acl(struct inode *inode, int type)
 
 	switch (type) {
 	case ACL_TYPE_ACCESS:
-		ea_name = SGI_ACL_FILE;
+		xfs_name_init(&name, SGI_ACL_FILE);
 		break;
 	case ACL_TYPE_DEFAULT:
-		ea_name = SGI_ACL_DEFAULT;
+		xfs_name_init(&name, SGI_ACL_DEFAULT);
 		break;
 	default:
 		BUG();
@@ -145,9 +145,8 @@ xfs_get_acl(struct inode *inode, int type)
 	 * go out to the disk.
 	 */
 	len = XFS_ACL_MAX_SIZE(ip->i_mount);
-	error = xfs_attr_get(ip, ea_name, strlen(ea_name),
-				(unsigned char **)&xfs_acl, &len,
-				ATTR_ALLOC | ATTR_ROOT);
+	error = xfs_attr_get(ip, &name, (unsigned char **)&xfs_acl, &len,
+			     ATTR_ALLOC | ATTR_ROOT);
 	if (error) {
 		/*
 		 * If the attribute doesn't exist make sure we have a negative
@@ -167,17 +166,17 @@ int
 __xfs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
 {
 	struct xfs_inode *ip = XFS_I(inode);
-	unsigned char *ea_name;
+	struct xfs_name name;
 	int error;
 
 	switch (type) {
 	case ACL_TYPE_ACCESS:
-		ea_name = SGI_ACL_FILE;
+		xfs_name_init(&name, SGI_ACL_FILE);
 		break;
 	case ACL_TYPE_DEFAULT:
 		if (!S_ISDIR(inode->i_mode))
 			return acl ? -EACCES : 0;
-		ea_name = SGI_ACL_DEFAULT;
+		xfs_name_init(&name, SGI_ACL_DEFAULT);
 		break;
 	default:
 		return -EINVAL;
@@ -197,17 +196,15 @@ __xfs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
 		len -= sizeof(struct xfs_acl_entry) *
 			 (XFS_ACL_MAX_ENTRIES(ip->i_mount) - acl->a_count);
 
-		error = xfs_attr_set(ip, ea_name, strlen(ea_name),
-				     (unsigned char *)xfs_acl, len, ATTR_ROOT);
+		error = xfs_attr_set(ip, &name, (unsigned char *)xfs_acl, len,
+				     ATTR_ROOT);
 
 		kmem_free(xfs_acl);
 	} else {
 		/*
 		 * A NULL ACL argument means we want to remove the ACL.
 		 */
-		error = xfs_attr_remove(ip, ea_name,
-					strlen(ea_name),
-					ATTR_ROOT);
+		error = xfs_attr_remove(ip, &name, ATTR_ROOT);
 
 		/*
 		 * If the attribute didn't exist to start with that's fine.
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index d42de92..28c07c9 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -357,7 +357,9 @@ xfs_attrmulti_attr_get(
 {
 	unsigned char		*kbuf;
 	int			error = -EFAULT;
-	size_t			namelen;
+	struct xfs_name		xname;
+
+	xfs_name_init(&xname, name);
 
 	if (*len > XFS_XATTR_SIZE_MAX)
 		return -EINVAL;
@@ -365,9 +367,7 @@ xfs_attrmulti_attr_get(
 	if (!kbuf)
 		return -ENOMEM;
 
-	namelen = strlen(name);
-	error = xfs_attr_get(XFS_I(inode), name, namelen, &kbuf, (int *)len,
-			     flags);
+	error = xfs_attr_get(XFS_I(inode), &xname, &kbuf, (int *)len, flags);
 	if (error)
 		goto out_kfree;
 
@@ -389,7 +389,9 @@ xfs_attrmulti_attr_set(
 {
 	unsigned char		*kbuf;
 	int			error;
-	size_t			namelen;
+	struct xfs_name		xname;
+
+	xfs_name_init(&xname, name);
 
 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 		return -EPERM;
@@ -400,8 +402,7 @@ xfs_attrmulti_attr_set(
 	if (IS_ERR(kbuf))
 		return PTR_ERR(kbuf);
 
-	namelen = strlen(name);
-	error = xfs_attr_set(XFS_I(inode), name, namelen, kbuf, len, flags);
+	error = xfs_attr_set(XFS_I(inode), &xname, kbuf, len, flags);
 	if (!error)
 		xfs_forget_acl(inode, name, flags);
 	kfree(kbuf);
@@ -415,12 +416,14 @@ xfs_attrmulti_attr_remove(
 	uint32_t		flags)
 {
 	int			error;
-	size_t			namelen;
+	struct xfs_name		xname;
+
+	xfs_name_init(&xname, name);
 
 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 		return -EPERM;
-	namelen = strlen(name);
-	error = xfs_attr_remove(XFS_I(inode), name, namelen, flags);
+
+	error = xfs_attr_remove(XFS_I(inode), &xname, flags);
 	if (!error)
 		xfs_forget_acl(inode, name, flags);
 	return error;
diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index 81f2f93..e85bbf5 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -48,11 +48,11 @@ xfs_initxattrs(
 	const struct xattr	*xattr;
 	struct xfs_inode	*ip = XFS_I(inode);
 	int			error = 0;
+	struct xfs_name		name;
 
 	for (xattr = xattr_array; xattr->name != NULL; xattr++) {
-		error = xfs_attr_set(ip, xattr->name,
-				     strlen(xattr->name),
-				     xattr->value, xattr->value_len,
+		xfs_name_init(&name, xattr->name);
+		error = xfs_attr_set(ip, &name, xattr->value, xattr->value_len,
 				     ATTR_SECURE);
 		if (error < 0)
 			break;
diff --git a/fs/xfs/xfs_xattr.c b/fs/xfs/xfs_xattr.c
index b0fedb5..74133a5 100644
--- a/fs/xfs/xfs_xattr.c
+++ b/fs/xfs/xfs_xattr.c
@@ -21,10 +21,12 @@ static int
 xfs_xattr_get(const struct xattr_handler *handler, struct dentry *unused,
 		struct inode *inode, const char *name, void *value, size_t size)
 {
-	int xflags = handler->flags;
-	struct xfs_inode *ip = XFS_I(inode);
-	int error, asize = size;
-	size_t namelen = strlen(name);
+	int			xflags = handler->flags;
+	struct xfs_inode	*ip = XFS_I(inode);
+	int			error, asize = size;
+	struct xfs_name		xname;
+
+	xfs_name_init(&xname, name);
 
 	/* Convert Linux syscall to XFS internal ATTR flags */
 	if (!size) {
@@ -32,8 +34,8 @@ xfs_xattr_get(const struct xattr_handler *handler, struct dentry *unused,
 		value = NULL;
 	}
 
-	error = xfs_attr_get(ip, name, namelen, (unsigned char **)&value,
-			     &asize, xflags);
+	error = xfs_attr_get(ip, &xname, (unsigned char **)&value, &asize,
+			     xflags);
 	if (error)
 		return error;
 	return asize;
@@ -69,7 +71,9 @@ xfs_xattr_set(const struct xattr_handler *handler, struct dentry *unused,
 	int			xflags = handler->flags;
 	struct xfs_inode	*ip = XFS_I(inode);
 	int			error;
-	size_t			namelen = strlen(name);
+	struct xfs_name		xname;
+
+	xfs_name_init(&xname, name);
 
 	/* Convert Linux syscall to XFS internal ATTR flags */
 	if (flags & XATTR_CREATE)
@@ -77,11 +81,11 @@ xfs_xattr_set(const struct xattr_handler *handler, struct dentry *unused,
 	if (flags & XATTR_REPLACE)
 		xflags |= ATTR_REPLACE;
 
-	if (value)
-		error = xfs_attr_set(ip, name, namelen, (void *)value, size,
-				xflags);
-	else
-		error = xfs_attr_remove(ip, name, namelen, xflags);
+        if (value)
+               error = xfs_attr_set(ip, &xname, (void *)value, size, xflags);
+        else
+               error = xfs_attr_remove(ip, &xname, xflags);
+
 	if (!error)
 		xfs_forget_acl(inode, name, xflags);
 
-- 
2.7.4

[PATCH v6 02/16] xfs: Embed struct xfs_name in xfs_da_args

[Allison Collins]

This patch embeds an xfs_name in xfs_da_args, replacing the name,
namelen, and flags members.  This helps to clean up the xfs_da_args
structure and make it more uniform with the new xfs_name parameter being
passed around.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
---
 fs/xfs/libxfs/xfs_attr.c        |  37 +++++++-------
 fs/xfs/libxfs/xfs_attr_leaf.c   | 104 +++++++++++++++++++++-------------------
 fs/xfs/libxfs/xfs_attr_remote.c |   2 +-
 fs/xfs/libxfs/xfs_da_btree.c    |   6 ++-
 fs/xfs/libxfs/xfs_da_btree.h    |   4 +-
 fs/xfs/libxfs/xfs_dir2.c        |  18 +++----
 fs/xfs/libxfs/xfs_dir2_block.c  |   6 +--
 fs/xfs/libxfs/xfs_dir2_leaf.c   |   6 +--
 fs/xfs/libxfs/xfs_dir2_node.c   |   8 ++--
 fs/xfs/libxfs/xfs_dir2_sf.c     |  30 ++++++------
 fs/xfs/scrub/attr.c             |  12 ++---
 fs/xfs/xfs_trace.h              |  20 ++++----
 12 files changed, 130 insertions(+), 123 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index 6717f47..9acdb23 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -72,13 +72,12 @@ xfs_attr_args_init(
 	args->geo = dp->i_mount->m_attr_geo;
 	args->whichfork = XFS_ATTR_FORK;
 	args->dp = dp;
-	args->flags = flags;
-	args->name = name->name;
-	args->namelen = name->len;
-	if (args->namelen >= MAXNAMELEN)
+	memcpy(&args->name, name, sizeof(struct xfs_name));
+	args->name.type = flags;
+	if (args->name.len >= MAXNAMELEN)
 		return -EFAULT;		/* match IRIX behaviour */
 
-	args->hashval = xfs_da_hashname(args->name, args->namelen);
+	args->hashval = xfs_da_hashname(args->name.name, args->name.len);
 	return 0;
 }
 
@@ -236,7 +235,7 @@ xfs_attr_try_sf_addname(
 	 * Commit the shortform mods, and we're done.
 	 * NOTE: this is also the error path (EEXIST, etc).
 	 */
-	if (!error && (args->flags & ATTR_KERNOTIME) == 0)
+	if (!error && (args->name.type & ATTR_KERNOTIME) == 0)
 		xfs_trans_ichgtime(args->trans, dp, XFS_ICHGTIME_CHG);
 
 	if (mp->m_flags & XFS_MOUNT_WSYNC)
@@ -357,6 +356,9 @@ xfs_attr_set(
 	if (error)
 		return error;
 
+	/* Use name now stored in args */
+	name = &args.name;
+
 	args.value = value;
 	args.valuelen = valuelen;
 	args.op_flags = XFS_DA_OP_ADDNAME | XFS_DA_OP_OKNOENT;
@@ -372,7 +374,7 @@ xfs_attr_set(
 	 */
 	if (XFS_IFORK_Q(dp) == 0) {
 		int sf_size = sizeof(xfs_attr_sf_hdr_t) +
-			XFS_ATTR_SF_ENTSIZE_BYNAME(args.namelen, valuelen);
+			XFS_ATTR_SF_ENTSIZE_BYNAME(name->len, valuelen);
 
 		error = xfs_bmap_add_attrfork(dp, sf_size, rsvd);
 		if (error)
@@ -457,6 +459,9 @@ xfs_attr_remove(
 	if (error)
 		return error;
 
+	/* Use name now stored in args */
+	name = &args.name;
+
 	/*
 	 * we have no control over the attribute names that userspace passes us
 	 * to remove, so we have to allow the name lookup prior to attribute
@@ -532,10 +537,10 @@ xfs_attr_shortform_addname(xfs_da_args_t *args)
 	trace_xfs_attr_sf_addname(args);
 
 	retval = xfs_attr_shortform_lookup(args);
-	if ((args->flags & ATTR_REPLACE) && (retval == -ENOATTR)) {
+	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
 		return retval;
 	} else if (retval == -EEXIST) {
-		if (args->flags & ATTR_CREATE)
+		if (args->name.type & ATTR_CREATE)
 			return retval;
 		retval = xfs_attr_shortform_remove(args);
 		if (retval)
@@ -545,15 +550,15 @@ xfs_attr_shortform_addname(xfs_da_args_t *args)
 		 * that the leaf format add routine won't trip over the attr
 		 * not being around.
 		 */
-		args->flags &= ~ATTR_REPLACE;
+		args->name.type &= ~ATTR_REPLACE;
 	}
 
-	if (args->namelen >= XFS_ATTR_SF_ENTSIZE_MAX ||
+	if (args->name.len >= XFS_ATTR_SF_ENTSIZE_MAX ||
 	    args->valuelen >= XFS_ATTR_SF_ENTSIZE_MAX)
 		return -ENOSPC;
 
 	newsize = XFS_ATTR_SF_TOTSIZE(args->dp);
-	newsize += XFS_ATTR_SF_ENTSIZE_BYNAME(args->namelen, args->valuelen);
+	newsize += XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
 
 	forkoff = xfs_attr_shortform_bytesfit(args->dp, newsize);
 	if (!forkoff)
@@ -598,11 +603,11 @@ xfs_attr_leaf_addname(
 	 * the given flags produce an error or call for an atomic rename.
 	 */
 	retval = xfs_attr3_leaf_lookup_int(bp, args);
-	if ((args->flags & ATTR_REPLACE) && (retval == -ENOATTR)) {
+	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
 		xfs_trans_brelse(args->trans, bp);
 		return retval;
 	} else if (retval == -EEXIST) {
-		if (args->flags & ATTR_CREATE) {	/* pure create op */
+		if (args->name.type & ATTR_CREATE) {	/* pure create op */
 			xfs_trans_brelse(args->trans, bp);
 			return retval;
 		}
@@ -872,10 +877,10 @@ xfs_attr_node_addname(
 		goto out;
 	blk = &state->path.blk[ state->path.active-1 ];
 	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
-	if ((args->flags & ATTR_REPLACE) && (retval == -ENOATTR)) {
+	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
 		goto out;
 	} else if (retval == -EEXIST) {
-		if (args->flags & ATTR_CREATE)
+		if (args->name.type & ATTR_CREATE)
 			goto out;
 
 		trace_xfs_attr_node_replace(args);
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
index fed537a..cb5ef66 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.c
+++ b/fs/xfs/libxfs/xfs_attr_leaf.c
@@ -464,7 +464,7 @@ xfs_attr_copy_value(
 	/*
 	 * No copy if all we have to do is get the length
 	 */
-	if (args->flags & ATTR_KERNOVAL) {
+	if (args->name.type & ATTR_KERNOVAL) {
 		args->valuelen = valuelen;
 		return 0;
 	}
@@ -679,27 +679,27 @@ xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff)
 	sfe = &sf->list[0];
 	for (i = 0; i < sf->hdr.count; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
 #ifdef DEBUG
-		if (sfe->namelen != args->namelen)
+		if (sfe->namelen != args->name.len)
 			continue;
-		if (memcmp(args->name, sfe->nameval, args->namelen) != 0)
+		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
 			continue;
-		if (!xfs_attr_namesp_match(args->flags, sfe->flags))
+		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
 			continue;
 		ASSERT(0);
 #endif
 	}
 
 	offset = (char *)sfe - (char *)sf;
-	size = XFS_ATTR_SF_ENTSIZE_BYNAME(args->namelen, args->valuelen);
+	size = XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
 	xfs_idata_realloc(dp, size, XFS_ATTR_FORK);
 	sf = (xfs_attr_shortform_t *)ifp->if_u1.if_data;
 	sfe = (xfs_attr_sf_entry_t *)((char *)sf + offset);
 
-	sfe->namelen = args->namelen;
+	sfe->namelen = args->name.len;
 	sfe->valuelen = args->valuelen;
-	sfe->flags = XFS_ATTR_NSP_ARGS_TO_ONDISK(args->flags);
-	memcpy(sfe->nameval, args->name, args->namelen);
-	memcpy(&sfe->nameval[args->namelen], args->value, args->valuelen);
+	sfe->flags = XFS_ATTR_NSP_ARGS_TO_ONDISK(args->name.type);
+	memcpy(sfe->nameval, args->name.name, args->name.len);
+	memcpy(&sfe->nameval[args->name.len], args->value, args->valuelen);
 	sf->hdr.count++;
 	be16_add_cpu(&sf->hdr.totsize, size);
 	xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE | XFS_ILOG_ADATA);
@@ -749,11 +749,11 @@ xfs_attr_shortform_remove(xfs_da_args_t *args)
 	for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe),
 					base += size, i++) {
 		size = XFS_ATTR_SF_ENTSIZE(sfe);
-		if (sfe->namelen != args->namelen)
+		if (sfe->namelen != args->name.len)
 			continue;
-		if (memcmp(sfe->nameval, args->name, args->namelen) != 0)
+		if (memcmp(sfe->nameval, args->name.name, args->name.len) != 0)
 			continue;
-		if (!xfs_attr_namesp_match(args->flags, sfe->flags))
+		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
 			continue;
 		break;
 	}
@@ -816,11 +816,11 @@ xfs_attr_shortform_lookup(xfs_da_args_t *args)
 	sfe = &sf->list[0];
 	for (i = 0; i < sf->hdr.count;
 				sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
-		if (sfe->namelen != args->namelen)
+		if (sfe->namelen != args->name.len)
 			continue;
-		if (memcmp(args->name, sfe->nameval, args->namelen) != 0)
+		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
 			continue;
-		if (!xfs_attr_namesp_match(args->flags, sfe->flags))
+		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
 			continue;
 		return -EEXIST;
 	}
@@ -847,13 +847,13 @@ xfs_attr_shortform_getvalue(
 	sfe = &sf->list[0];
 	for (i = 0; i < sf->hdr.count;
 				sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
-		if (sfe->namelen != args->namelen)
+		if (sfe->namelen != args->name.len)
 			continue;
-		if (memcmp(args->name, sfe->nameval, args->namelen) != 0)
+		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
 			continue;
-		if (!xfs_attr_namesp_match(args->flags, sfe->flags))
+		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
 			continue;
-		return xfs_attr_copy_value(args, &sfe->nameval[args->namelen],
+		return xfs_attr_copy_value(args, &sfe->nameval[args->name.len],
 						sfe->valuelen);
 	}
 	return -ENOATTR;
@@ -912,13 +912,13 @@ xfs_attr_shortform_to_leaf(
 
 	sfe = &sf->list[0];
 	for (i = 0; i < sf->hdr.count; i++) {
-		nargs.name = sfe->nameval;
-		nargs.namelen = sfe->namelen;
-		nargs.value = &sfe->nameval[nargs.namelen];
+		nargs.name.name = sfe->nameval;
+		nargs.name.len = sfe->namelen;
+		nargs.value = &sfe->nameval[nargs.name.len];
 		nargs.valuelen = sfe->valuelen;
 		nargs.hashval = xfs_da_hashname(sfe->nameval,
 						sfe->namelen);
-		nargs.flags = XFS_ATTR_NSP_ONDISK_TO_ARGS(sfe->flags);
+		nargs.name.type = XFS_ATTR_NSP_ONDISK_TO_ARGS(sfe->flags);
 		error = xfs_attr3_leaf_lookup_int(bp, &nargs); /* set a->index */
 		ASSERT(error == -ENOATTR);
 		error = xfs_attr3_leaf_add(bp, &nargs);
@@ -1119,12 +1119,12 @@ xfs_attr3_leaf_to_shortform(
 			continue;
 		ASSERT(entry->flags & XFS_ATTR_LOCAL);
 		name_loc = xfs_attr3_leaf_name_local(leaf, i);
-		nargs.name = name_loc->nameval;
-		nargs.namelen = name_loc->namelen;
-		nargs.value = &name_loc->nameval[nargs.namelen];
+		nargs.name.name = name_loc->nameval;
+		nargs.name.len = name_loc->namelen;
+		nargs.value = &name_loc->nameval[nargs.name.len];
 		nargs.valuelen = be16_to_cpu(name_loc->valuelen);
 		nargs.hashval = be32_to_cpu(entry->hashval);
-		nargs.flags = XFS_ATTR_NSP_ONDISK_TO_ARGS(entry->flags);
+		nargs.name.type = XFS_ATTR_NSP_ONDISK_TO_ARGS(entry->flags);
 		xfs_attr_shortform_add(&nargs, forkoff);
 	}
 	error = 0;
@@ -1450,7 +1450,7 @@ xfs_attr3_leaf_add_work(
 				     ichdr->freemap[mapindex].size);
 	entry->hashval = cpu_to_be32(args->hashval);
 	entry->flags = tmp ? XFS_ATTR_LOCAL : 0;
-	entry->flags |= XFS_ATTR_NSP_ARGS_TO_ONDISK(args->flags);
+	entry->flags |= XFS_ATTR_NSP_ARGS_TO_ONDISK(args->name.type);
 	if (args->op_flags & XFS_DA_OP_RENAME) {
 		entry->flags |= XFS_ATTR_INCOMPLETE;
 		if ((args->blkno2 == args->blkno) &&
@@ -1474,15 +1474,16 @@ xfs_attr3_leaf_add_work(
 	 */
 	if (entry->flags & XFS_ATTR_LOCAL) {
 		name_loc = xfs_attr3_leaf_name_local(leaf, args->index);
-		name_loc->namelen = args->namelen;
+		name_loc->namelen = args->name.len;
 		name_loc->valuelen = cpu_to_be16(args->valuelen);
-		memcpy((char *)name_loc->nameval, args->name, args->namelen);
-		memcpy((char *)&name_loc->nameval[args->namelen], args->value,
+		memcpy((char *)name_loc->nameval, args->name.name,
+		       args->name.len);
+		memcpy((char *)&name_loc->nameval[args->name.len], args->value,
 				   be16_to_cpu(name_loc->valuelen));
 	} else {
 		name_rmt = xfs_attr3_leaf_name_remote(leaf, args->index);
-		name_rmt->namelen = args->namelen;
-		memcpy((char *)name_rmt->name, args->name, args->namelen);
+		name_rmt->namelen = args->name.len;
+		memcpy((char *)name_rmt->name, args->name.name, args->name.len);
 		entry->flags |= XFS_ATTR_INCOMPLETE;
 		/* just in case */
 		name_rmt->valuelen = 0;
@@ -2409,23 +2410,25 @@ xfs_attr3_leaf_lookup_int(
 		}
 		if (entry->flags & XFS_ATTR_LOCAL) {
 			name_loc = xfs_attr3_leaf_name_local(leaf, probe);
-			if (name_loc->namelen != args->namelen)
+			if (name_loc->namelen != args->name.len)
 				continue;
-			if (memcmp(args->name, name_loc->nameval,
-							args->namelen) != 0)
+			if (memcmp(args->name.name, name_loc->nameval,
+							args->name.len) != 0)
 				continue;
-			if (!xfs_attr_namesp_match(args->flags, entry->flags))
+			if (!xfs_attr_namesp_match(args->name.type,
+						   entry->flags))
 				continue;
 			args->index = probe;
 			return -EEXIST;
 		} else {
 			name_rmt = xfs_attr3_leaf_name_remote(leaf, probe);
-			if (name_rmt->namelen != args->namelen)
+			if (name_rmt->namelen != args->name.len)
 				continue;
-			if (memcmp(args->name, name_rmt->name,
-							args->namelen) != 0)
+			if (memcmp(args->name.name, name_rmt->name,
+							args->name.len) != 0)
 				continue;
-			if (!xfs_attr_namesp_match(args->flags, entry->flags))
+			if (!xfs_attr_namesp_match(args->name.type,
+						   entry->flags))
 				continue;
 			args->index = probe;
 			args->rmtvaluelen = be32_to_cpu(name_rmt->valuelen);
@@ -2467,16 +2470,17 @@ xfs_attr3_leaf_getvalue(
 	entry = &xfs_attr3_leaf_entryp(leaf)[args->index];
 	if (entry->flags & XFS_ATTR_LOCAL) {
 		name_loc = xfs_attr3_leaf_name_local(leaf, args->index);
-		ASSERT(name_loc->namelen == args->namelen);
-		ASSERT(memcmp(args->name, name_loc->nameval, args->namelen) == 0);
+		ASSERT(name_loc->namelen == args->name.len);
+		ASSERT(memcmp(args->name.name, name_loc->nameval,
+			      args->name.len) == 0);
 		return xfs_attr_copy_value(args,
-					&name_loc->nameval[args->namelen],
+					&name_loc->nameval[args->name.len],
 					be16_to_cpu(name_loc->valuelen));
 	}
 
 	name_rmt = xfs_attr3_leaf_name_remote(leaf, args->index);
-	ASSERT(name_rmt->namelen == args->namelen);
-	ASSERT(memcmp(args->name, name_rmt->name, args->namelen) == 0);
+	ASSERT(name_rmt->namelen == args->name.len);
+	ASSERT(memcmp(args->name.name, name_rmt->name, args->name.len) == 0);
 	args->rmtvaluelen = be32_to_cpu(name_rmt->valuelen);
 	args->rmtblkno = be32_to_cpu(name_rmt->valueblk);
 	args->rmtblkcnt = xfs_attr3_rmt_blocks(args->dp->i_mount,
@@ -2692,7 +2696,7 @@ xfs_attr_leaf_newentsize(
 {
 	int			size;
 
-	size = xfs_attr_leaf_entsize_local(args->namelen, args->valuelen);
+	size = xfs_attr_leaf_entsize_local(args->name.len, args->valuelen);
 	if (size < xfs_attr_leaf_entsize_local_max(args->geo->blksize)) {
 		if (local)
 			*local = 1;
@@ -2700,7 +2704,7 @@ xfs_attr_leaf_newentsize(
 	}
 	if (local)
 		*local = 0;
-	return xfs_attr_leaf_entsize_remote(args->namelen);
+	return xfs_attr_leaf_entsize_remote(args->name.len);
 }
 
 
@@ -2754,8 +2758,8 @@ xfs_attr3_leaf_clearflag(
 		name = (char *)name_rmt->name;
 	}
 	ASSERT(be32_to_cpu(entry->hashval) == args->hashval);
-	ASSERT(namelen == args->namelen);
-	ASSERT(memcmp(name, args->name, namelen) == 0);
+	ASSERT(namelen == args->name.len);
+	ASSERT(memcmp(name, args->name.name, namelen) == 0);
 #endif /* DEBUG */
 
 	entry->flags &= ~XFS_ATTR_INCOMPLETE;
diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
index a6ef5df..3e725c3 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.c
+++ b/fs/xfs/libxfs/xfs_attr_remote.c
@@ -380,7 +380,7 @@ xfs_attr_rmtval_get(
 
 	trace_xfs_attr_rmtval_get(args);
 
-	ASSERT(!(args->flags & ATTR_KERNOVAL));
+	ASSERT(!(args->name.type & ATTR_KERNOVAL));
 	ASSERT(args->rmtvaluelen == args->valuelen);
 
 	valuelen = args->rmtvaluelen;
diff --git a/fs/xfs/libxfs/xfs_da_btree.c b/fs/xfs/libxfs/xfs_da_btree.c
index 8c3eafe..da7e1af 100644
--- a/fs/xfs/libxfs/xfs_da_btree.c
+++ b/fs/xfs/libxfs/xfs_da_btree.c
@@ -2125,8 +2125,10 @@ xfs_da_compname(
 	const unsigned char *name,
 	int		len)
 {
-	return (args->namelen == len && memcmp(args->name, name, len) == 0) ?
-					XFS_CMP_EXACT : XFS_CMP_DIFFERENT;
+	if (args->name.len == len && !memcmp(args->name.name, name, len))
+		return XFS_CMP_EXACT;
+
+	return XFS_CMP_DIFFERENT;
 }
 
 int
diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
index 0f4fbb0..14f1be3 100644
--- a/fs/xfs/libxfs/xfs_da_btree.h
+++ b/fs/xfs/libxfs/xfs_da_btree.h
@@ -54,12 +54,10 @@ enum xfs_dacmp {
  */
 typedef struct xfs_da_args {
 	struct xfs_da_geometry *geo;	/* da block geometry */
-	const uint8_t		*name;		/* string (maybe not NULL terminated) */
-	int		namelen;	/* length of string (maybe no NULL) */
+	struct xfs_name	name;		/* name, length and argument  flags*/
 	uint8_t		filetype;	/* filetype of inode for directories */
 	uint8_t		*value;		/* set of bytes (maybe contain NULLs) */
 	int		valuelen;	/* length of value */
-	int		flags;		/* argument flags (eg: ATTR_NOCREATE) */
 	xfs_dahash_t	hashval;	/* hash value of name */
 	xfs_ino_t	inumber;	/* input/output inode number */
 	struct xfs_inode *dp;		/* directory inode to manipulate */
diff --git a/fs/xfs/libxfs/xfs_dir2.c b/fs/xfs/libxfs/xfs_dir2.c
index dd6fcaa..3aadddc 100644
--- a/fs/xfs/libxfs/xfs_dir2.c
+++ b/fs/xfs/libxfs/xfs_dir2.c
@@ -74,14 +74,14 @@ xfs_ascii_ci_compname(
 	enum xfs_dacmp		result;
 	int			i;
 
-	if (args->namelen != len)
+	if (args->name.len != len)
 		return XFS_CMP_DIFFERENT;
 
 	result = XFS_CMP_EXACT;
 	for (i = 0; i < len; i++) {
-		if (args->name[i] == name[i])
+		if (args->name.name[i] == name[i])
 			continue;
-		if (tolower(args->name[i]) != tolower(name[i]))
+		if (tolower(args->name.name[i]) != tolower(name[i]))
 			return XFS_CMP_DIFFERENT;
 		result = XFS_CMP_CASE;
 	}
@@ -265,8 +265,7 @@ xfs_dir_createname(
 		return -ENOMEM;
 
 	args->geo = dp->i_mount->m_dir_geo;
-	args->name = name->name;
-	args->namelen = name->len;
+	memcpy(&args->name, name, sizeof(struct xfs_name));
 	args->filetype = name->type;
 	args->hashval = xfs_dir2_hashname(dp->i_mount, name);
 	args->inumber = inum;
@@ -361,8 +360,7 @@ xfs_dir_lookup(
 	 */
 	args = kmem_zalloc(sizeof(*args), KM_NOFS);
 	args->geo = dp->i_mount->m_dir_geo;
-	args->name = name->name;
-	args->namelen = name->len;
+	memcpy(&args->name, name, sizeof(struct xfs_name));
 	args->filetype = name->type;
 	args->hashval = xfs_dir2_hashname(dp->i_mount, name);
 	args->dp = dp;
@@ -433,8 +431,7 @@ xfs_dir_removename(
 		return -ENOMEM;
 
 	args->geo = dp->i_mount->m_dir_geo;
-	args->name = name->name;
-	args->namelen = name->len;
+	memcpy(&args->name, name, sizeof(struct xfs_name));
 	args->filetype = name->type;
 	args->hashval = xfs_dir2_hashname(dp->i_mount, name);
 	args->inumber = ino;
@@ -494,8 +491,7 @@ xfs_dir_replace(
 		return -ENOMEM;
 
 	args->geo = dp->i_mount->m_dir_geo;
-	args->name = name->name;
-	args->namelen = name->len;
+	memcpy(&args->name, name, sizeof(struct xfs_name));
 	args->filetype = name->type;
 	args->hashval = xfs_dir2_hashname(dp->i_mount, name);
 	args->inumber = inum;
diff --git a/fs/xfs/libxfs/xfs_dir2_block.c b/fs/xfs/libxfs/xfs_dir2_block.c
index d6ced59..592e47c 100644
--- a/fs/xfs/libxfs/xfs_dir2_block.c
+++ b/fs/xfs/libxfs/xfs_dir2_block.c
@@ -355,7 +355,7 @@ xfs_dir2_block_addname(
 	if (error)
 		return error;
 
-	len = xfs_dir2_data_entsize(dp->i_mount, args->namelen);
+	len = xfs_dir2_data_entsize(dp->i_mount, args->name.len);
 
 	/*
 	 * Set up pointers to parts of the block.
@@ -539,8 +539,8 @@ xfs_dir2_block_addname(
 	 * Create the new data entry.
 	 */
 	dep->inumber = cpu_to_be64(args->inumber);
-	dep->namelen = args->namelen;
-	memcpy(dep->name, args->name, args->namelen);
+	dep->namelen = args->name.len;
+	memcpy(dep->name, args->name.name, args->name.len);
 	xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
 	tagp = xfs_dir2_data_entry_tag_p(dp->i_mount, dep);
 	*tagp = cpu_to_be16((char *)dep - (char *)hdr);
diff --git a/fs/xfs/libxfs/xfs_dir2_leaf.c b/fs/xfs/libxfs/xfs_dir2_leaf.c
index a131b52..24a7fda 100644
--- a/fs/xfs/libxfs/xfs_dir2_leaf.c
+++ b/fs/xfs/libxfs/xfs_dir2_leaf.c
@@ -653,7 +653,7 @@ xfs_dir2_leaf_addname(
 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
 	ents = leafhdr.ents;
 	bestsp = xfs_dir2_leaf_bests_p(ltp);
-	length = xfs_dir2_data_entsize(dp->i_mount, args->namelen);
+	length = xfs_dir2_data_entsize(dp->i_mount, args->name.len);
 
 	/*
 	 * See if there are any entries with the same hash value
@@ -856,8 +856,8 @@ xfs_dir2_leaf_addname(
 	 */
 	dep = (xfs_dir2_data_entry_t *)dup;
 	dep->inumber = cpu_to_be64(args->inumber);
-	dep->namelen = args->namelen;
-	memcpy(dep->name, args->name, dep->namelen);
+	dep->namelen = args->name.len;
+	memcpy(dep->name, args->name.name, dep->namelen);
 	xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
 	tagp = xfs_dir2_data_entry_tag_p(dp->i_mount, dep);
 	*tagp = cpu_to_be16((char *)dep - (char *)hdr);
diff --git a/fs/xfs/libxfs/xfs_dir2_node.c b/fs/xfs/libxfs/xfs_dir2_node.c
index a0cc5e2..3c74efc 100644
--- a/fs/xfs/libxfs/xfs_dir2_node.c
+++ b/fs/xfs/libxfs/xfs_dir2_node.c
@@ -666,7 +666,7 @@ xfs_dir2_leafn_lookup_for_addname(
 		ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC) ||
 		       free->hdr.magic == cpu_to_be32(XFS_DIR3_FREE_MAGIC));
 	}
-	length = xfs_dir2_data_entsize(mp, args->namelen);
+	length = xfs_dir2_data_entsize(mp, args->name.len);
 	/*
 	 * Loop over leaf entries with the right hash value.
 	 */
@@ -1911,7 +1911,7 @@ xfs_dir2_node_addname_int(
 	int			needscan = 0;	/* need to rescan data frees */
 	__be16			*tagp;		/* data entry tag pointer */
 
-	length = xfs_dir2_data_entsize(dp->i_mount, args->namelen);
+	length = xfs_dir2_data_entsize(dp->i_mount, args->name.len);
 	error = xfs_dir2_node_find_freeblk(args, fblk, &dbno, &fbp, &freehdr,
 					   &findex, length);
 	if (error)
@@ -1966,8 +1966,8 @@ xfs_dir2_node_addname_int(
 	/* Fill in the new entry and log it. */
 	dep = (xfs_dir2_data_entry_t *)dup;
 	dep->inumber = cpu_to_be64(args->inumber);
-	dep->namelen = args->namelen;
-	memcpy(dep->name, args->name, dep->namelen);
+	dep->namelen = args->name.len;
+	memcpy(dep->name, args->name.name, dep->namelen);
 	xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
 	tagp = xfs_dir2_data_entry_tag_p(dp->i_mount, dep);
 	*tagp = cpu_to_be16((char *)dep - (char *)hdr);
diff --git a/fs/xfs/libxfs/xfs_dir2_sf.c b/fs/xfs/libxfs/xfs_dir2_sf.c
index 7b7f6fb..058c526 100644
--- a/fs/xfs/libxfs/xfs_dir2_sf.c
+++ b/fs/xfs/libxfs/xfs_dir2_sf.c
@@ -387,7 +387,7 @@ xfs_dir2_sf_addname(
 	/*
 	 * Compute entry (and change in) size.
 	 */
-	incr_isize = xfs_dir2_sf_entsize(dp->i_mount, sfp, args->namelen);
+	incr_isize = xfs_dir2_sf_entsize(dp->i_mount, sfp, args->name.len);
 	objchange = 0;
 
 	/*
@@ -470,7 +470,7 @@ xfs_dir2_sf_addname_easy(
 	/*
 	 * Grow the in-inode space.
 	 */
-	xfs_idata_realloc(dp, xfs_dir2_sf_entsize(mp, sfp, args->namelen),
+	xfs_idata_realloc(dp, xfs_dir2_sf_entsize(mp, sfp, args->name.len),
 			  XFS_DATA_FORK);
 	/*
 	 * Need to set up again due to realloc of the inode data.
@@ -480,9 +480,9 @@ xfs_dir2_sf_addname_easy(
 	/*
 	 * Fill in the new entry.
 	 */
-	sfep->namelen = args->namelen;
+	sfep->namelen = args->name.len;
 	xfs_dir2_sf_put_offset(sfep, offset);
-	memcpy(sfep->name, args->name, sfep->namelen);
+	memcpy(sfep->name, args->name.name, sfep->namelen);
 	xfs_dir2_sf_put_ino(mp, sfp, sfep, args->inumber);
 	xfs_dir2_sf_put_ftype(mp, sfep, args->filetype);
 
@@ -540,7 +540,7 @@ xfs_dir2_sf_addname_hard(
 	 */
 	for (offset = args->geo->data_first_offset,
 	      oldsfep = xfs_dir2_sf_firstentry(oldsfp),
-	      add_datasize = xfs_dir2_data_entsize(mp, args->namelen),
+	      add_datasize = xfs_dir2_data_entsize(mp, args->name.len),
 	      eof = (char *)oldsfep == &buf[old_isize];
 	     !eof;
 	     offset = new_offset + xfs_dir2_data_entsize(mp, oldsfep->namelen),
@@ -570,9 +570,9 @@ xfs_dir2_sf_addname_hard(
 	/*
 	 * Fill in the new entry, and update the header counts.
 	 */
-	sfep->namelen = args->namelen;
+	sfep->namelen = args->name.len;
 	xfs_dir2_sf_put_offset(sfep, offset);
-	memcpy(sfep->name, args->name, sfep->namelen);
+	memcpy(sfep->name, args->name.name, sfep->namelen);
 	xfs_dir2_sf_put_ino(mp, sfp, sfep, args->inumber);
 	xfs_dir2_sf_put_ftype(mp, sfep, args->filetype);
 	sfp->count++;
@@ -615,7 +615,7 @@ xfs_dir2_sf_addname_pick(
 	int			used;		/* data bytes used */
 
 	sfp = (xfs_dir2_sf_hdr_t *)dp->i_df.if_u1.if_data;
-	size = xfs_dir2_data_entsize(mp, args->namelen);
+	size = xfs_dir2_data_entsize(mp, args->name.len);
 	offset = args->geo->data_first_offset;
 	sfep = xfs_dir2_sf_firstentry(sfp);
 	holefit = 0;
@@ -887,7 +887,7 @@ xfs_dir2_sf_lookup(
 	/*
 	 * Special case for .
 	 */
-	if (args->namelen == 1 && args->name[0] == '.') {
+	if (args->name.len == 1 && args->name.name[0] == '.') {
 		args->inumber = dp->i_ino;
 		args->cmpresult = XFS_CMP_EXACT;
 		args->filetype = XFS_DIR3_FT_DIR;
@@ -896,8 +896,8 @@ xfs_dir2_sf_lookup(
 	/*
 	 * Special case for ..
 	 */
-	if (args->namelen == 2 &&
-	    args->name[0] == '.' && args->name[1] == '.') {
+	if (args->name.len == 2 &&
+	    args->name.name[0] == '.' && args->name.name[1] == '.') {
 		args->inumber = xfs_dir2_sf_get_parent_ino(sfp);
 		args->cmpresult = XFS_CMP_EXACT;
 		args->filetype = XFS_DIR3_FT_DIR;
@@ -984,7 +984,7 @@ xfs_dir2_sf_removename(
 	 * Calculate sizes.
 	 */
 	byteoff = (int)((char *)sfep - (char *)sfp);
-	entsize = xfs_dir2_sf_entsize(mp, sfp, args->namelen);
+	entsize = xfs_dir2_sf_entsize(mp, sfp, args->name.len);
 	newsize = oldsize - entsize;
 	/*
 	 * Copy the part if any after the removed entry, sliding it down.
@@ -1085,12 +1085,12 @@ xfs_dir2_sf_replace(
 	} else
 		i8elevated = 0;
 
-	ASSERT(args->namelen != 1 || args->name[0] != '.');
+	ASSERT(args->name.len != 1 || args->name.name[0] != '.');
 	/*
 	 * Replace ..'s entry.
 	 */
-	if (args->namelen == 2 &&
-	    args->name[0] == '.' && args->name[1] == '.') {
+	if (args->name.len == 2 &&
+	    args->name.name[0] == '.' && args->name.name[1] == '.') {
 		ino = xfs_dir2_sf_get_parent_ino(sfp);
 		ASSERT(args->inumber != ino);
 		xfs_dir2_sf_put_parent_ino(sfp, args->inumber);
diff --git a/fs/xfs/scrub/attr.c b/fs/xfs/scrub/attr.c
index d9f0dd4..d4a9fe4 100644
--- a/fs/xfs/scrub/attr.c
+++ b/fs/xfs/scrub/attr.c
@@ -147,17 +147,17 @@ xchk_xattr_listent(
 		return;
 	}
 
-	args.flags = ATTR_KERNOTIME;
+	args.name.type = ATTR_KERNOTIME;
 	if (flags & XFS_ATTR_ROOT)
-		args.flags |= ATTR_ROOT;
+		args.name.type |= ATTR_ROOT;
 	else if (flags & XFS_ATTR_SECURE)
-		args.flags |= ATTR_SECURE;
+		args.name.type |= ATTR_SECURE;
 	args.geo = context->dp->i_mount->m_attr_geo;
 	args.whichfork = XFS_ATTR_FORK;
 	args.dp = context->dp;
-	args.name = name;
-	args.namelen = namelen;
-	args.hashval = xfs_da_hashname(args.name, args.namelen);
+	args.name.name = name;
+	args.name.len = namelen;
+	args.hashval = xfs_da_hashname(args.name.name, args.name.len);
 	args.trans = context->tp;
 	args.value = xchk_xattr_valuebuf(sx->sc);
 	args.valuelen = valuelen;
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index a86be7f..159b8af 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -1633,7 +1633,7 @@ DECLARE_EVENT_CLASS(xfs_da_class,
 	TP_STRUCT__entry(
 		__field(dev_t, dev)
 		__field(xfs_ino_t, ino)
-		__dynamic_array(char, name, args->namelen)
+		__dynamic_array(char, name, args->name.len)
 		__field(int, namelen)
 		__field(xfs_dahash_t, hashval)
 		__field(xfs_ino_t, inumber)
@@ -1642,9 +1642,10 @@ DECLARE_EVENT_CLASS(xfs_da_class,
 	TP_fast_assign(
 		__entry->dev = VFS_I(args->dp)->i_sb->s_dev;
 		__entry->ino = args->dp->i_ino;
-		if (args->namelen)
-			memcpy(__get_str(name), args->name, args->namelen);
-		__entry->namelen = args->namelen;
+		if (args->name.len)
+			memcpy(__get_str(name), args->name.name,
+			       args->name.len);
+		__entry->namelen = args->name.len;
 		__entry->hashval = args->hashval;
 		__entry->inumber = args->inumber;
 		__entry->op_flags = args->op_flags;
@@ -1697,7 +1698,7 @@ DECLARE_EVENT_CLASS(xfs_attr_class,
 	TP_STRUCT__entry(
 		__field(dev_t, dev)
 		__field(xfs_ino_t, ino)
-		__dynamic_array(char, name, args->namelen)
+		__dynamic_array(char, name, args->name.len)
 		__field(int, namelen)
 		__field(int, valuelen)
 		__field(xfs_dahash_t, hashval)
@@ -1707,12 +1708,13 @@ DECLARE_EVENT_CLASS(xfs_attr_class,
 	TP_fast_assign(
 		__entry->dev = VFS_I(args->dp)->i_sb->s_dev;
 		__entry->ino = args->dp->i_ino;
-		if (args->namelen)
-			memcpy(__get_str(name), args->name, args->namelen);
-		__entry->namelen = args->namelen;
+		if (args->name.len)
+			memcpy(__get_str(name), args->name.name,
+			       args->name.len);
+		__entry->namelen = args->name.len;
 		__entry->valuelen = args->valuelen;
 		__entry->hashval = args->hashval;
-		__entry->flags = args->flags;
+		__entry->flags = args->name.type;
 		__entry->op_flags = args->op_flags;
 	),
 	TP_printk("dev %d:%d ino 0x%llx name %.*s namelen %d valuelen %d "
-- 
2.7.4

[PATCH v6 03/16] xfs: Add xfs_has_attr and subroutines

[Allison Collins]

From: Allison Henderson <allison.henderson@oracle.com>

This patch adds a new functions to check for the existence of an
attribute.  Subroutines are also added to handle the cases of leaf
blocks, nodes or shortform.  Common code that appears in existing attr
add and remove functions have been factored out to help reduce the
appearance of duplicated code.  We will need these routines later for
delayed attributes since delayed operations cannot return error codes.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c      | 171 ++++++++++++++++++++++++++++--------------
 fs/xfs/libxfs/xfs_attr.h      |   1 +
 fs/xfs/libxfs/xfs_attr_leaf.c | 111 +++++++++++++++++----------
 fs/xfs/libxfs/xfs_attr_leaf.h |   3 +
 4 files changed, 188 insertions(+), 98 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index 9acdb23..a2507b9 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -46,6 +46,7 @@ STATIC int xfs_attr_shortform_addname(xfs_da_args_t *args);
 STATIC int xfs_attr_leaf_get(xfs_da_args_t *args);
 STATIC int xfs_attr_leaf_addname(xfs_da_args_t *args);
 STATIC int xfs_attr_leaf_removename(xfs_da_args_t *args);
+STATIC int xfs_attr_leaf_hasname(struct xfs_da_args *args, struct xfs_buf **bp);
 
 /*
  * Internal routines when attribute list is more than one block.
@@ -53,6 +54,8 @@ STATIC int xfs_attr_leaf_removename(xfs_da_args_t *args);
 STATIC int xfs_attr_node_get(xfs_da_args_t *args);
 STATIC int xfs_attr_node_addname(xfs_da_args_t *args);
 STATIC int xfs_attr_node_removename(xfs_da_args_t *args);
+STATIC int xfs_attr_node_hasname(xfs_da_args_t *args,
+				 struct xfs_da_state **state);
 STATIC int xfs_attr_fillstate(xfs_da_state_t *state);
 STATIC int xfs_attr_refillstate(xfs_da_state_t *state);
 
@@ -310,6 +313,37 @@ xfs_attr_set_args(
 }
 
 /*
+ * Return EEXIST if attr is found, or ENOATTR if not
+ */
+int
+xfs_has_attr(
+	struct xfs_da_args      *args)
+{
+	struct xfs_inode	*dp = args->dp;
+	struct xfs_buf		*bp = NULL;
+	int			error;
+
+	if (!xfs_inode_hasattr(dp))
+		return -ENOATTR;
+
+	if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
+		ASSERT(dp->i_afp->if_flags & XFS_IFINLINE);
+		return xfs_attr_sf_findname(args, NULL, NULL);
+	}
+
+	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
+		error = xfs_attr_leaf_hasname(args, &bp);
+
+		if (bp)
+			xfs_trans_brelse(args->trans, bp);
+
+		return error;
+	}
+
+	return xfs_attr_node_hasname(args, NULL);
+}
+
+/*
  * Remove the attribute specified in @args.
  */
 int
@@ -583,26 +617,20 @@ STATIC int
 xfs_attr_leaf_addname(
 	struct xfs_da_args	*args)
 {
-	struct xfs_inode	*dp;
 	struct xfs_buf		*bp;
 	int			retval, error, forkoff;
+	struct xfs_inode	*dp = args->dp;
 
 	trace_xfs_attr_leaf_addname(args);
 
 	/*
-	 * Read the (only) block in the attribute list in.
-	 */
-	dp = args->dp;
-	args->blkno = 0;
-	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
-	if (error)
-		return error;
-
-	/*
 	 * Look up the given attribute in the leaf block.  Figure out if
 	 * the given flags produce an error or call for an atomic rename.
 	 */
-	retval = xfs_attr3_leaf_lookup_int(bp, args);
+	retval = xfs_attr_leaf_hasname(args, &bp);
+	if (retval != -ENOATTR && retval != -EEXIST)
+		return retval;
+
 	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
 		xfs_trans_brelse(args->trans, bp);
 		return retval;
@@ -754,6 +782,27 @@ xfs_attr_leaf_addname(
 }
 
 /*
+ * Return EEXIST if attr is found, or ENOATTR if not
+ */
+STATIC int
+xfs_attr_leaf_hasname(
+	struct xfs_da_args      *args,
+	struct xfs_buf		**bp)
+{
+	int                     error = 0;
+
+	error = xfs_attr3_leaf_read(args->trans, args->dp, 0, bp);
+	if (error)
+		return error;
+
+	error = xfs_attr3_leaf_lookup_int(*bp, args);
+	if (error != -ENOATTR && error != -EEXIST)
+		xfs_trans_brelse(args->trans, *bp);
+
+	return error;
+}
+
+/*
  * Remove a name from the leaf attribute list structure
  *
  * This leaf block cannot have a "remote" value, we only call this routine
@@ -773,12 +822,11 @@ xfs_attr_leaf_removename(
 	 * Remove the attribute.
 	 */
 	dp = args->dp;
-	args->blkno = 0;
-	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
-	if (error)
+
+	error = xfs_attr_leaf_hasname(args, &bp);
+	if (error != -ENOATTR && error != -EEXIST)
 		return error;
 
-	error = xfs_attr3_leaf_lookup_int(bp, args);
 	if (error == -ENOATTR) {
 		xfs_trans_brelse(args->trans, bp);
 		return error;
@@ -817,12 +865,10 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
 
 	trace_xfs_attr_leaf_get(args);
 
-	args->blkno = 0;
-	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
-	if (error)
+	error = xfs_attr_leaf_hasname(args, &bp);
+	if (error != -ENOATTR && error != -EEXIST)
 		return error;
 
-	error = xfs_attr3_leaf_lookup_int(bp, args);
 	if (error != -EEXIST)  {
 		xfs_trans_brelse(args->trans, bp);
 		return error;
@@ -832,6 +878,41 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
 	return error;
 }
 
+/*
+ * Return EEXIST if attr is found, or ENOATTR if not
+ * statep: If not null is set to point at the found state.  Caller will
+ *         be responsible for freeing the state in this case.
+ */
+STATIC int
+xfs_attr_node_hasname(
+	struct xfs_da_args	*args,
+	struct xfs_da_state	**statep)
+{
+	struct xfs_da_state	*state;
+	int			retval, error;
+
+	state = xfs_da_state_alloc();
+	state->args = args;
+	state->mp = args->dp->i_mount;
+
+	if (statep != NULL)
+		*statep = NULL;
+
+	/*
+	 * Search to see if name exists, and get back a pointer to it.
+	 */
+	error = xfs_da3_node_lookup_int(state, &retval);
+	if (error == 0) {
+		if (statep != NULL)
+			*statep = state;
+		return retval;
+	}
+
+	xfs_da_state_free(state);
+
+	return error;
+}
+
 /*========================================================================
  * External routines when attribute list size > geo->blksize
  *========================================================================*/
@@ -864,20 +945,17 @@ xfs_attr_node_addname(
 	dp = args->dp;
 	mp = dp->i_mount;
 restart:
-	state = xfs_da_state_alloc();
-	state->args = args;
-	state->mp = mp;
-
 	/*
 	 * Search to see if name already exists, and get back a pointer
 	 * to where it should go.
 	 */
-	error = xfs_da3_node_lookup_int(state, &retval);
-	if (error)
+	retval = xfs_attr_node_hasname(args, &state);
+	if (retval != -ENOATTR)
 		goto out;
+
 	blk = &state->path.blk[ state->path.active-1 ];
 	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
-	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
+	if (args->name.type & ATTR_REPLACE) {
 		goto out;
 	} else if (retval == -EEXIST) {
 		if (args->name.type & ATTR_CREATE)
@@ -1079,29 +1157,15 @@ xfs_attr_node_removename(
 {
 	struct xfs_da_state	*state;
 	struct xfs_da_state_blk	*blk;
-	struct xfs_inode	*dp;
 	struct xfs_buf		*bp;
 	int			retval, error, forkoff;
+	struct xfs_inode	*dp = args->dp;
 
 	trace_xfs_attr_node_removename(args);
 
-	/*
-	 * Tie a string around our finger to remind us where we are.
-	 */
-	dp = args->dp;
-	state = xfs_da_state_alloc();
-	state->args = args;
-	state->mp = dp->i_mount;
-
-	/*
-	 * Search to see if name exists, and get back a pointer to it.
-	 */
-	error = xfs_da3_node_lookup_int(state, &retval);
-	if (error || (retval != -EEXIST)) {
-		if (error == 0)
-			error = retval;
+	error = xfs_attr_node_hasname(args, &state);
+	if (error != -EEXIST)
 		goto out;
-	}
 
 	/*
 	 * If there is an out-of-line value, de-allocate the blocks.
@@ -1196,7 +1260,8 @@ xfs_attr_node_removename(
 	error = 0;
 
 out:
-	xfs_da_state_free(state);
+	if (state)
+		xfs_da_state_free(state);
 	return error;
 }
 
@@ -1316,31 +1381,23 @@ xfs_attr_node_get(xfs_da_args_t *args)
 {
 	xfs_da_state_t *state;
 	xfs_da_state_blk_t *blk;
-	int error, retval;
+	int error;
 	int i;
 
 	trace_xfs_attr_node_get(args);
 
-	state = xfs_da_state_alloc();
-	state->args = args;
-	state->mp = args->dp->i_mount;
-
 	/*
 	 * Search to see if name exists, and get back a pointer to it.
 	 */
-	error = xfs_da3_node_lookup_int(state, &retval);
-	if (error) {
-		retval = error;
-		goto out_release;
-	}
-	if (retval != -EEXIST)
+	error = xfs_attr_node_hasname(args, &state);
+	if (error != -EEXIST)
 		goto out_release;
 
 	/*
 	 * Get the value, local or "remote"
 	 */
 	blk = &state->path.blk[state->path.active - 1];
-	retval = xfs_attr3_leaf_getvalue(blk->bp, args);
+	error = xfs_attr3_leaf_getvalue(blk->bp, args);
 
 	/*
 	 * If not in a transaction, we have to release all the buffers.
@@ -1352,7 +1409,7 @@ xfs_attr_node_get(xfs_da_args_t *args)
 	}
 
 	xfs_da_state_free(state);
-	return retval;
+	return error;
 }
 
 /* Returns true if the attribute entry name is valid. */
diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
index 35043db..ce7b039 100644
--- a/fs/xfs/libxfs/xfs_attr.h
+++ b/fs/xfs/libxfs/xfs_attr.h
@@ -153,6 +153,7 @@ int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
 		 unsigned char *value, int valuelen, int flags);
 int xfs_attr_set_args(struct xfs_da_args *args);
 int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
+int xfs_has_attr(struct xfs_da_args *args);
 int xfs_attr_remove_args(struct xfs_da_args *args);
 int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
 		  int flags, struct attrlist_cursor_kern *cursor);
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
index cb5ef66..9d6b68c 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.c
+++ b/fs/xfs/libxfs/xfs_attr_leaf.c
@@ -654,18 +654,66 @@ xfs_attr_shortform_create(xfs_da_args_t *args)
 }
 
 /*
+ * Return -EEXIST if attr is found, or -ENOATTR if not
+ * args:  args containing attribute name and namelen
+ * sfep:  If not null, pointer will be set to the last attr entry found on
+	  -EEXIST.  On -ENOATTR pointer is left at the last entry in the list
+ * basep: If not null, pointer is set to the byte offset of the entry in the
+ *	  list on -EEXIST.  On -ENOATTR, pointer is left at the byte offset of
+ *	  the last entry in the list
+ */
+int
+xfs_attr_sf_findname(
+	struct xfs_da_args	 *args,
+	struct xfs_attr_sf_entry **sfep,
+	unsigned int		 *basep)
+{
+	struct xfs_attr_shortform *sf;
+	struct xfs_attr_sf_entry *sfe;
+	unsigned int		base = sizeof(struct xfs_attr_sf_hdr);
+	int			size = 0;
+	int			end;
+	int			i;
+
+	sf = (struct xfs_attr_shortform *)args->dp->i_afp->if_u1.if_data;
+	sfe = &sf->list[0];
+	end = sf->hdr.count;
+	for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe),
+			base += size, i++) {
+		size = XFS_ATTR_SF_ENTSIZE(sfe);
+		if (sfe->namelen != args->name.len)
+			continue;
+		if (memcmp(sfe->nameval, args->name.name, args->name.len) != 0)
+			continue;
+		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
+			continue;
+		break;
+	}
+
+	if (sfep != NULL)
+		*sfep = sfe;
+
+	if (basep != NULL)
+		*basep = base;
+
+	if (i == end)
+		return -ENOATTR;
+	return -EEXIST;
+}
+
+/*
  * Add a name/value pair to the shortform attribute list.
  * Overflow from the inode has already been checked for.
  */
 void
-xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff)
+xfs_attr_shortform_add(struct xfs_da_args *args, int forkoff)
 {
-	xfs_attr_shortform_t *sf;
-	xfs_attr_sf_entry_t *sfe;
-	int i, offset, size;
-	xfs_mount_t *mp;
-	xfs_inode_t *dp;
-	struct xfs_ifork *ifp;
+	struct xfs_attr_shortform	*sf;
+	struct xfs_attr_sf_entry	*sfe;
+	int				offset, size, error;
+	struct xfs_mount		*mp;
+	struct xfs_inode		*dp;
+	struct xfs_ifork		*ifp;
 
 	trace_xfs_attr_sf_add(args);
 
@@ -676,18 +724,8 @@ xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff)
 	ifp = dp->i_afp;
 	ASSERT(ifp->if_flags & XFS_IFINLINE);
 	sf = (xfs_attr_shortform_t *)ifp->if_u1.if_data;
-	sfe = &sf->list[0];
-	for (i = 0; i < sf->hdr.count; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
-#ifdef DEBUG
-		if (sfe->namelen != args->name.len)
-			continue;
-		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
-			continue;
-		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
-			continue;
-		ASSERT(0);
-#endif
-	}
+	error = xfs_attr_sf_findname(args, &sfe, NULL);
+	ASSERT(error != -EEXIST);
 
 	offset = (char *)sfe - (char *)sf;
 	size = XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
@@ -730,35 +768,26 @@ xfs_attr_fork_remove(
  * Remove an attribute from the shortform attribute list structure.
  */
 int
-xfs_attr_shortform_remove(xfs_da_args_t *args)
+xfs_attr_shortform_remove(struct xfs_da_args *args)
 {
-	xfs_attr_shortform_t *sf;
-	xfs_attr_sf_entry_t *sfe;
-	int base, size=0, end, totsize, i;
-	xfs_mount_t *mp;
-	xfs_inode_t *dp;
+	struct xfs_attr_shortform	*sf;
+	struct xfs_attr_sf_entry	*sfe;
+	int				size = 0, end, totsize;
+	unsigned int			base;
+	struct xfs_mount		*mp;
+	struct xfs_inode		*dp;
+	int				error;
 
 	trace_xfs_attr_sf_remove(args);
 
 	dp = args->dp;
 	mp = dp->i_mount;
-	base = sizeof(xfs_attr_sf_hdr_t);
 	sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data;
-	sfe = &sf->list[0];
-	end = sf->hdr.count;
-	for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe),
-					base += size, i++) {
-		size = XFS_ATTR_SF_ENTSIZE(sfe);
-		if (sfe->namelen != args->name.len)
-			continue;
-		if (memcmp(sfe->nameval, args->name.name, args->name.len) != 0)
-			continue;
-		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
-			continue;
-		break;
-	}
-	if (i == end)
-		return -ENOATTR;
+
+	error = xfs_attr_sf_findname(args, &sfe, &base);
+	if (error != -EEXIST)
+		return error;
+	size = XFS_ATTR_SF_ENTSIZE(sfe);
 
 	/*
 	 * Fix up the attribute fork data, covering the hole
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.h b/fs/xfs/libxfs/xfs_attr_leaf.h
index f4a188e..f6165ff 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.h
+++ b/fs/xfs/libxfs/xfs_attr_leaf.h
@@ -62,6 +62,9 @@ int	xfs_attr_shortform_getvalue(struct xfs_da_args *args);
 int	xfs_attr_shortform_to_leaf(struct xfs_da_args *args,
 			struct xfs_buf **leaf_bp);
 int	xfs_attr_shortform_remove(struct xfs_da_args *args);
+int	xfs_attr_sf_findname(struct xfs_da_args *args,
+			     struct xfs_attr_sf_entry **sfep,
+			     unsigned int *basep);
 int	xfs_attr_shortform_allfit(struct xfs_buf *bp, struct xfs_inode *dp);
 int	xfs_attr_shortform_bytesfit(struct xfs_inode *dp, int bytes);
 xfs_failaddr_t xfs_attr_shortform_verify(struct xfs_inode *ip);
-- 
2.7.4

[PATCH v6 04/16] xfs: Factor out new helper functions xfs_attr_rmtval_set

[Allison Collins]

Break xfs_attr_rmtval_set into two helper functions
xfs_attr_rmt_find_hole and xfs_attr_rmtval_set_value.
xfs_attr_rmtval_set rolls the transaction between the
helpers, but delayed operations cannot.  We will use the helpers later
when constructing new delayed attribute routines.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_attr_remote.c | 149 +++++++++++++++++++++++++---------------
 1 file changed, 92 insertions(+), 57 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
index 3e725c3..c2ab80d 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.c
+++ b/fs/xfs/libxfs/xfs_attr_remote.c
@@ -425,32 +425,23 @@ xfs_attr_rmtval_get(
 }
 
 /*
- * Write the value associated with an attribute into the out-of-line buffer
- * that we have defined for it.
+ * Find a "hole" in the attribute address space large enough for us to drop the
+ * new attribute's value into
  */
-int
-xfs_attr_rmtval_set(
+STATIC int
+xfs_attr_rmt_find_hole(
 	struct xfs_da_args	*args)
 {
 	struct xfs_inode	*dp = args->dp;
 	struct xfs_mount	*mp = dp->i_mount;
-	struct xfs_bmbt_irec	map;
-	xfs_dablk_t		lblkno;
-	xfs_fileoff_t		lfileoff = 0;
-	uint8_t			*src = args->value;
-	int			blkcnt;
-	int			valuelen;
-	int			nmap;
 	int			error;
-	int			offset = 0;
-
-	trace_xfs_attr_rmtval_set(args);
+	int			blkcnt;
+	xfs_fileoff_t		lfileoff = 0;
 
 	/*
-	 * Find a "hole" in the attribute address space large enough for
-	 * us to drop the new attribute's value into. Because CRC enable
-	 * attributes have headers, we can't just do a straight byte to FSB
-	 * conversion and have to take the header space into account.
+	 * Because CRC enable attributes have headers, we can't just do a
+	 * straight byte to FSB conversion and have to take the header space
+	 * into account.
 	 */
 	blkcnt = xfs_attr3_rmt_blocks(mp, args->rmtvaluelen);
 	error = xfs_bmap_first_unused(args->trans, args->dp, blkcnt, &lfileoff,
@@ -458,48 +449,26 @@ xfs_attr_rmtval_set(
 	if (error)
 		return error;
 
-	args->rmtblkno = lblkno = (xfs_dablk_t)lfileoff;
+	args->rmtblkno = (xfs_dablk_t)lfileoff;
 	args->rmtblkcnt = blkcnt;
 
-	/*
-	 * Roll through the "value", allocating blocks on disk as required.
-	 */
-	while (blkcnt > 0) {
-		/*
-		 * Allocate a single extent, up to the size of the value.
-		 *
-		 * Note that we have to consider this a data allocation as we
-		 * write the remote attribute without logging the contents.
-		 * Hence we must ensure that we aren't using blocks that are on
-		 * the busy list so that we don't overwrite blocks which have
-		 * recently been freed but their transactions are not yet
-		 * committed to disk. If we overwrite the contents of a busy
-		 * extent and then crash then the block may not contain the
-		 * correct metadata after log recovery occurs.
-		 */
-		nmap = 1;
-		error = xfs_bmapi_write(args->trans, dp, (xfs_fileoff_t)lblkno,
-				  blkcnt, XFS_BMAPI_ATTRFORK, args->total, &map,
-				  &nmap);
-		if (error)
-			return error;
-		error = xfs_defer_finish(&args->trans);
-		if (error)
-			return error;
-
-		ASSERT(nmap == 1);
-		ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
-		       (map.br_startblock != HOLESTARTBLOCK));
-		lblkno += map.br_blockcount;
-		blkcnt -= map.br_blockcount;
+	return 0;
+}
 
-		/*
-		 * Start the next trans in the chain.
-		 */
-		error = xfs_trans_roll_inode(&args->trans, dp);
-		if (error)
-			return error;
-	}
+STATIC int
+xfs_attr_rmtval_set_value(
+	struct xfs_da_args	*args)
+{
+	struct xfs_inode	*dp = args->dp;
+	struct xfs_mount	*mp = dp->i_mount;
+	struct xfs_bmbt_irec	map;
+	xfs_dablk_t		lblkno;
+	uint8_t			*src = args->value;
+	int			blkcnt;
+	int			valuelen;
+	int			nmap;
+	int			error;
+	int			offset = 0;
 
 	/*
 	 * Roll through the "value", copying the attribute value to the
@@ -553,6 +522,72 @@ xfs_attr_rmtval_set(
 }
 
 /*
+ * Write the value associated with an attribute into the out-of-line buffer
+ * that we have defined for it.
+ */
+int
+xfs_attr_rmtval_set(
+	struct xfs_da_args	*args)
+{
+	struct xfs_inode	*dp = args->dp;
+	struct xfs_bmbt_irec	map;
+	xfs_dablk_t		lblkno;
+	int			blkcnt;
+	int			nmap;
+	int			error;
+
+	trace_xfs_attr_rmtval_set(args);
+
+	error = xfs_attr_rmt_find_hole(args);
+	if (error)
+		return error;
+
+	blkcnt = args->rmtblkcnt;
+	lblkno = (xfs_dablk_t)args->rmtblkno;
+	/*
+	 * Roll through the "value", allocating blocks on disk as required.
+	 */
+	while (blkcnt > 0) {
+		/*
+		 * Allocate a single extent, up to the size of the value.
+		 *
+		 * Note that we have to consider this a data allocation as we
+		 * write the remote attribute without logging the contents.
+		 * Hence we must ensure that we aren't using blocks that are on
+		 * the busy list so that we don't overwrite blocks which have
+		 * recently been freed but their transactions are not yet
+		 * committed to disk. If we overwrite the contents of a busy
+		 * extent and then crash then the block may not contain the
+		 * correct metadata after log recovery occurs.
+		 */
+		nmap = 1;
+		error = xfs_bmapi_write(args->trans, dp, (xfs_fileoff_t)lblkno,
+				  blkcnt, XFS_BMAPI_ATTRFORK, args->total, &map,
+				  &nmap);
+		if (error)
+			return error;
+		error = xfs_defer_finish(&args->trans);
+		if (error)
+			return error;
+
+		ASSERT(nmap == 1);
+		ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
+		       (map.br_startblock != HOLESTARTBLOCK));
+		lblkno += map.br_blockcount;
+		blkcnt -= map.br_blockcount;
+
+		/*
+		 * Start the next trans in the chain.
+		 */
+		error = xfs_trans_roll_inode(&args->trans, dp);
+		if (error)
+			return error;
+	}
+
+	return xfs_attr_rmtval_set_value(args);
+}
+
+/*
  * Remove the value associated with an attribute by deleting the
  * out-of-line buffer that it is stored on.
  */
-- 
2.7.4

[PATCH v6 05/16] xfs: Factor out trans handling in xfs_attr3_leaf_flipflags

[Allison Collins]

Since delayed operations cannot roll transactions, factor up the
transaction handling into the calling function

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c      | 14 ++++++++++++++
 fs/xfs/libxfs/xfs_attr_leaf.c |  7 +------
 2 files changed, 15 insertions(+), 6 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index a2507b9..b0ec25b 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -727,6 +727,13 @@ xfs_attr_leaf_addname(
 		error = xfs_attr3_leaf_flipflags(args);
 		if (error)
 			return error;
+		/*
+		 * Commit the flag value change and start the next trans in
+		 * series.
+		 */
+		error = xfs_trans_roll_inode(&args->trans, args->dp);
+		if (error)
+			return error;
 
 		/*
 		 * Dismantle the "old" attribute/value pair by removing
@@ -1069,6 +1076,13 @@ xfs_attr_node_addname(
 		error = xfs_attr3_leaf_flipflags(args);
 		if (error)
 			goto out;
+		/*
+		 * Commit the flag value change and start the next trans in
+		 * series
+		 */
+		error = xfs_trans_roll_inode(&args->trans, args->dp);
+		if (error)
+			goto out;
 
 		/*
 		 * Dismantle the "old" attribute/value pair by removing
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
index 9d6b68c..d691509 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.c
+++ b/fs/xfs/libxfs/xfs_attr_leaf.c
@@ -2973,10 +2973,5 @@ xfs_attr3_leaf_flipflags(
 			 XFS_DA_LOGRANGE(leaf2, name_rmt, sizeof(*name_rmt)));
 	}
 
-	/*
-	 * Commit the flag value change and start the next trans in series.
-	 */
-	error = xfs_trans_roll_inode(&args->trans, args->dp);
-
-	return error;
+	return 0;
 }
-- 
2.7.4

[PATCH v6 06/16] xfs: Factor out xfs_attr_leaf_addname helper

[Allison Collins]

Factor out new helper function xfs_attr_leaf_try_add. Because new
delayed attribute routines cannot roll transactions, we carve off the
parts of xfs_attr_leaf_addname that we can use, and move the commit into
the calling function.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
---
 fs/xfs/libxfs/xfs_attr.c | 83 ++++++++++++++++++++++++++++--------------------
 1 file changed, 49 insertions(+), 34 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index b0ec25b..9ed7e94 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -305,10 +305,30 @@ xfs_attr_set_args(
 		}
 	}
 
-	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK))
+	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
 		error = xfs_attr_leaf_addname(args);
-	else
-		error = xfs_attr_node_addname(args);
+		if (error != -ENOSPC)
+			return error;
+
+		/*
+		 * Commit that transaction so that the node_addname()
+		 * call can manage its own transactions.
+		 */
+		error = xfs_defer_finish(&args->trans);
+		if (error)
+			return error;
+
+		/*
+		 * Commit the current trans (including the inode) and
+		 * start a new one.
+		 */
+		error = xfs_trans_roll_inode(&args->trans, dp);
+		if (error)
+			return error;
+
+	}
+
+	error = xfs_attr_node_addname(args);
 	return error;
 }
 
@@ -607,21 +627,12 @@ xfs_attr_shortform_addname(xfs_da_args_t *args)
  * External routines when attribute list is one block
  *========================================================================*/
 
-/*
- * Add a name to the leaf attribute list structure
- *
- * This leaf block cannot have a "remote" value, we only call this routine
- * if bmap_one_block() says there is only one block (ie: no remote blks).
- */
 STATIC int
-xfs_attr_leaf_addname(
-	struct xfs_da_args	*args)
+xfs_attr_leaf_try_add(
+	struct xfs_da_args	*args,
+	struct xfs_buf		*bp)
 {
-	struct xfs_buf		*bp;
-	int			retval, error, forkoff;
-	struct xfs_inode	*dp = args->dp;
-
-	trace_xfs_attr_leaf_addname(args);
+	int			retval, error;
 
 	/*
 	 * Look up the given attribute in the leaf block.  Figure out if
@@ -667,31 +678,35 @@ xfs_attr_leaf_addname(
 	retval = xfs_attr3_leaf_add(bp, args);
 	if (retval == -ENOSPC) {
 		/*
-		 * Promote the attribute list to the Btree format, then
-		 * Commit that transaction so that the node_addname() call
-		 * can manage its own transactions.
+		 * Promote the attribute list to the Btree format.
+		 * Unless an error occurs, retain the -ENOSPC retval
 		 */
 		error = xfs_attr3_leaf_to_node(args);
 		if (error)
 			return error;
-		error = xfs_defer_finish(&args->trans);
-		if (error)
-			return error;
+	}
+	return retval;
+}
 
-		/*
-		 * Commit the current trans (including the inode) and start
-		 * a new one.
-		 */
-		error = xfs_trans_roll_inode(&args->trans, dp);
-		if (error)
-			return error;
 
-		/*
-		 * Fob the whole rest of the problem off on the Btree code.
-		 */
-		error = xfs_attr_node_addname(args);
+/*
+ * Add a name to the leaf attribute list structure
+ *
+ * This leaf block cannot have a "remote" value, we only call this routine
+ * if bmap_one_block() says there is only one block (ie: no remote blks).
+ */
+STATIC int
+xfs_attr_leaf_addname(struct xfs_da_args	*args)
+{
+	int			error, forkoff;
+	struct xfs_buf		*bp = NULL;
+	struct xfs_inode	*dp = args->dp;
+
+	trace_xfs_attr_leaf_addname(args);
+
+	error = xfs_attr_leaf_try_add(args, bp);
+	if (error)
 		return error;
-	}
 
 	/*
 	 * Commit the transaction that added the attr name so that
-- 
2.7.4

[PATCH v6 07/16] xfs: Refactor xfs_attr_try_sf_addname

[Allison Collins]

To help pre-simplify xfs_attr_set_args, we need to hoist transacation
handling up, while modularizing the adjacent code down into helpers. In
this patch, hoist the commit in xfs_attr_try_sf_addname up into the
calling function, and also pull the attr list creation down.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c | 30 +++++++++++++++---------------
 1 file changed, 15 insertions(+), 15 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index 9ed7e94..c15361a 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -227,8 +227,13 @@ xfs_attr_try_sf_addname(
 	struct xfs_da_args	*args)
 {
 
-	struct xfs_mount	*mp = dp->i_mount;
-	int			error, error2;
+	int			error;
+
+	/*
+	 * Build initial attribute list (if required).
+	 */
+	if (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS)
+		xfs_attr_shortform_create(args);
 
 	error = xfs_attr_shortform_addname(args);
 	if (error == -ENOSPC)
@@ -241,12 +246,10 @@ xfs_attr_try_sf_addname(
 	if (!error && (args->name.type & ATTR_KERNOTIME) == 0)
 		xfs_trans_ichgtime(args->trans, dp, XFS_ICHGTIME_CHG);
 
-	if (mp->m_flags & XFS_MOUNT_WSYNC)
+	if (dp->i_mount->m_flags & XFS_MOUNT_WSYNC)
 		xfs_trans_set_sync(args->trans);
 
-	error2 = xfs_trans_commit(args->trans);
-	args->trans = NULL;
-	return error ? error : error2;
+	return error;
 }
 
 /*
@@ -258,7 +261,7 @@ xfs_attr_set_args(
 {
 	struct xfs_inode	*dp = args->dp;
 	struct xfs_buf          *leaf_bp = NULL;
-	int			error;
+	int			error, error2 = 0;
 
 	/*
 	 * If the attribute list is non-existent or a shortform list,
@@ -269,17 +272,14 @@ xfs_attr_set_args(
 	     dp->i_d.di_anextents == 0)) {
 
 		/*
-		 * Build initial attribute list (if required).
-		 */
-		if (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS)
-			xfs_attr_shortform_create(args);
-
-		/*
 		 * Try to add the attr to the attribute list in the inode.
 		 */
 		error = xfs_attr_try_sf_addname(dp, args);
-		if (error != -ENOSPC)
-			return error;
+		if (error != -ENOSPC) {
+			error2 = xfs_trans_commit(args->trans);
+			args->trans = NULL;
+			return error ? error : error2;
+		}
 
 		/*
 		 * It won't fit in the shortform, transform to a leaf block.
-- 
2.7.4

[PATCH v6 08/16] xfs: Factor out trans roll from xfs_attr3_leaf_setflag

[Allison Collins]

New delayed allocation routines cannot be handling transactions so
factor them up into the calling functions

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
---
 fs/xfs/libxfs/xfs_attr.c      | 5 +++++
 fs/xfs/libxfs/xfs_attr_leaf.c | 5 +----
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index c15361a..8ddea8d 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -1221,6 +1221,11 @@ xfs_attr_node_removename(
 		error = xfs_attr3_leaf_setflag(args);
 		if (error)
 			goto out;
+
+		error = xfs_trans_roll_inode(&args->trans, args->dp);
+		if (error)
+			goto out;
+
 		error = xfs_attr_rmtval_remove(args);
 		if (error)
 			goto out;
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
index d691509..67339f0 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.c
+++ b/fs/xfs/libxfs/xfs_attr_leaf.c
@@ -2855,10 +2855,7 @@ xfs_attr3_leaf_setflag(
 			 XFS_DA_LOGRANGE(leaf, name_rmt, sizeof(*name_rmt)));
 	}
 
-	/*
-	 * Commit the flag value change and start the next trans in series.
-	 */
-	return xfs_trans_roll_inode(&args->trans, args->dp);
+	return 0;
 }
 
 /*
-- 
2.7.4

[PATCH v6 09/16] xfs: Factor out xfs_attr_rmtval_invalidate

[Allison Collins]

Because new delayed attribute routines cannot roll transactions, we
carve off the parts of xfs_attr_rmtval_remove that we can use.  This
will help to reduce repetitive code later when we introduce delayed
attributes.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_attr_remote.c | 26 +++++++++++++++++++++-----
 fs/xfs/libxfs/xfs_attr_remote.h |  2 +-
 2 files changed, 22 insertions(+), 6 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
index c2ab80d..9b4fa2a 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.c
+++ b/fs/xfs/libxfs/xfs_attr_remote.c
@@ -592,16 +592,13 @@ xfs_attr_rmtval_set(
  * out-of-line buffer that it is stored on.
  */
 int
-xfs_attr_rmtval_remove(
+xfs_attr_rmtval_invalidate(
 	struct xfs_da_args	*args)
 {
 	struct xfs_mount	*mp = args->dp->i_mount;
 	xfs_dablk_t		lblkno;
 	int			blkcnt;
 	int			error;
-	int			done;
-
-	trace_xfs_attr_rmtval_remove(args);
 
 	/*
 	 * Roll through the "value", invalidating the attribute value's blocks.
@@ -643,13 +640,32 @@ xfs_attr_rmtval_remove(
 		lblkno += map.br_blockcount;
 		blkcnt -= map.br_blockcount;
 	}
+	return 0;
+}
 
+/*
+ * Remove the value associated with an attribute by deleting the
+ * out-of-line buffer that it is stored on.
+ */
+int
+xfs_attr_rmtval_remove(
+	struct xfs_da_args      *args)
+{
+	xfs_dablk_t		lblkno;
+	int			blkcnt;
+	int			error = 0;
+	int			done = 0;
+
+	trace_xfs_attr_rmtval_remove(args);
+
+	error = xfs_attr_rmtval_invalidate(args);
+	if (error)
+		return error;
 	/*
 	 * Keep de-allocating extents until the remote-value region is gone.
 	 */
 	lblkno = args->rmtblkno;
 	blkcnt = args->rmtblkcnt;
-	done = 0;
 	while (!done) {
 		error = xfs_bunmapi(args->trans, args->dp, lblkno, blkcnt,
 				    XFS_BMAPI_ATTRFORK, 1, &done);
diff --git a/fs/xfs/libxfs/xfs_attr_remote.h b/fs/xfs/libxfs/xfs_attr_remote.h
index 9d20b66..85f2593 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.h
+++ b/fs/xfs/libxfs/xfs_attr_remote.h
@@ -11,5 +11,5 @@ int xfs_attr3_rmt_blocks(struct xfs_mount *mp, int attrlen);
 int xfs_attr_rmtval_get(struct xfs_da_args *args);
 int xfs_attr_rmtval_set(struct xfs_da_args *args);
 int xfs_attr_rmtval_remove(struct xfs_da_args *args);
-
+int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
 #endif /* __XFS_ATTR_REMOTE_H__ */
-- 
2.7.4

[PATCH v6 10/16] xfs: Factor out trans roll in xfs_attr3_leaf_clearflag

[Allison Collins]

New delayed allocation routines cannot be handling transactions so
factor them out into the calling functions

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c      | 16 ++++++++++++++++
 fs/xfs/libxfs/xfs_attr_leaf.c |  5 +----
 2 files changed, 17 insertions(+), 4 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index 8ddea8d..a2673fe 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -799,6 +799,14 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
 		 * Added a "remote" value, just clear the incomplete flag.
 		 */
 		error = xfs_attr3_leaf_clearflag(args);
+		if (error)
+			return error;
+
+		/*
+		 * Commit the flag value change and start the next trans in
+		 * series.
+		 */
+		error = xfs_trans_roll_inode(&args->trans, args->dp);
 	}
 	return error;
 }
@@ -1162,6 +1170,14 @@ xfs_attr_node_addname(
 		error = xfs_attr3_leaf_clearflag(args);
 		if (error)
 			goto out;
+
+		 /*
+		  * Commit the flag value change and start the next trans in
+		  * series.
+		  */
+		error = xfs_trans_roll_inode(&args->trans, args->dp);
+		if (error)
+			goto out;
 	}
 	retval = error = 0;
 
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
index 67339f0..1742f0a 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.c
+++ b/fs/xfs/libxfs/xfs_attr_leaf.c
@@ -2804,10 +2804,7 @@ xfs_attr3_leaf_clearflag(
 			 XFS_DA_LOGRANGE(leaf, name_rmt, sizeof(*name_rmt)));
 	}
 
-	/*
-	 * Commit the flag value change and start the next trans in series.
-	 */
-	return xfs_trans_roll_inode(&args->trans, args->dp);
+	return 0;
 }
 
 /*
-- 
2.7.4

[PATCH v6 11/16] xfs: Check for -ENOATTR or -EEXIST

[Allison Collins]

Delayed operations cannot return error codes.  So we must check for
these conditions first before starting set or remove operations

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index a2673fe..e9d22c1 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -457,6 +457,14 @@ xfs_attr_set(
 		goto out_trans_cancel;
 
 	xfs_trans_ijoin(args.trans, dp, 0);
+
+	error = xfs_has_attr(&args);
+	if (error == -EEXIST && (name->type & ATTR_CREATE))
+		goto out_trans_cancel;
+
+	if (error == -ENOATTR && (name->type & ATTR_REPLACE))
+		goto out_trans_cancel;
+
 	error = xfs_attr_set_args(&args);
 	if (error)
 		goto out_trans_cancel;
@@ -545,6 +553,10 @@ xfs_attr_remove(
 	 */
 	xfs_trans_ijoin(args.trans, dp, 0);
 
+	error = xfs_has_attr(&args);
+	if (error != -EEXIST)
+		goto out;
+
 	error = xfs_attr_remove_args(&args);
 	if (error)
 		goto out;
-- 
2.7.4

[PATCH v6 12/16] xfs: Add helper function xfs_attr_init_unmapstate

[Allison Collins]

This patch helps to pre-simplify xfs_attr_node_removename by modularizing
the code around the transactions into helper functions.  This will make
the function easier to follow when we introduce delayed attributes.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c | 45 +++++++++++++++++++++++++++++++--------------
 1 file changed, 31 insertions(+), 14 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index e9d22c1..453ea59 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -1202,6 +1202,36 @@ xfs_attr_node_addname(
 }
 
 /*
+ * Set up the state for unmap and set the incomplete flag
+ */
+STATIC int
+xfs_attr_init_unmapstate(
+	struct xfs_da_args	*args,
+	struct xfs_da_state	*state)
+{
+	int error;
+
+	/*
+	 * Fill in disk block numbers in the state structure
+	 * so that we can get the buffers back after we commit
+	 * several transactions in the following calls.
+	 */
+	error = xfs_attr_fillstate(state);
+	if (error)
+		return error;
+
+	/*
+	 * Mark the attribute as INCOMPLETE
+	 */
+	error = xfs_attr3_leaf_setflag(args);
+	if (error)
+		return error;
+
+	return 0;
+}
+
+
+/*
  * Remove a name from a B-tree attribute list.
  *
  * This will involve walking down the Btree, and may involve joining
@@ -1233,20 +1263,7 @@ xfs_attr_node_removename(
 	ASSERT(blk->bp != NULL);
 	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
 	if (args->rmtblkno > 0) {
-		/*
-		 * Fill in disk block numbers in the state structure
-		 * so that we can get the buffers back after we commit
-		 * several transactions in the following calls.
-		 */
-		error = xfs_attr_fillstate(state);
-		if (error)
-			goto out;
-
-		/*
-		 * Mark the attribute as INCOMPLETE, then bunmapi() the
-		 * remote value.
-		 */
-		error = xfs_attr3_leaf_setflag(args);
+		error = xfs_attr_init_unmapstate(args, state);
 		if (error)
 			goto out;
 
-- 
2.7.4

[PATCH v6 13/16] xfs: Add helper function xfs_attr_rmtval_unmap

[Allison Collins]

This function is similar to xfs_attr_rmtval_remove, but adapted to
return. EAGAIN for new transactions. We will use this later when we
introduce delayed attributes

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr_remote.c | 27 +++++++++++++++++++++++++++
 fs/xfs/libxfs/xfs_attr_remote.h |  1 +
 2 files changed, 28 insertions(+)

diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
index 9b4fa2a..f2ee0b8 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.c
+++ b/fs/xfs/libxfs/xfs_attr_remote.c
@@ -684,3 +684,30 @@ xfs_attr_rmtval_remove(
 	}
 	return 0;
 }
+
+/*
+ * Unmap value blocks for this attr.  This is similar to
+ * xfs_attr_rmtval_remove, but adapted to to return EAGAIN for new transactions
+ */
+int
+xfs_attr_rmtval_unmap(
+	struct xfs_da_args	*args)
+{
+	int	error, done;
+
+	/*
+	 * Unmap value blocks for this attr.  This is similar to
+	 * xfs_attr_rmtval_remove, but open coded here to return EAGAIN
+	 * for new transactions
+	 */
+	error = xfs_bunmapi(args->trans, args->dp,
+		    args->rmtblkno, args->rmtblkcnt,
+		    XFS_BMAPI_ATTRFORK, 1, &done);
+	if (error)
+		return error;
+
+	if (!done)
+		return -EAGAIN;
+
+	return 0;
+}
diff --git a/fs/xfs/libxfs/xfs_attr_remote.h b/fs/xfs/libxfs/xfs_attr_remote.h
index 85f2593..7ab3770 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.h
+++ b/fs/xfs/libxfs/xfs_attr_remote.h
@@ -12,4 +12,5 @@ int xfs_attr_rmtval_get(struct xfs_da_args *args);
 int xfs_attr_rmtval_set(struct xfs_da_args *args);
 int xfs_attr_rmtval_remove(struct xfs_da_args *args);
 int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
+int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
 #endif /* __XFS_ATTR_REMOTE_H__ */
-- 
2.7.4

[PATCH v6 14/16] xfs: Simplify xfs_attr_set_args

[Allison Collins]

Delayed attribute mechanics make frequent use of goto statements.  We
can use this to further simplify xfs_attr_set_args.  In this patch we
introduce one of the gotos now to help pre-simplify the routine.  This
will help make proceeding patches in this area easier to read.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c | 71 +++++++++++++++++++++++++-----------------------
 1 file changed, 37 insertions(+), 34 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index 453ea59..90e0b2d 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -264,47 +264,50 @@ xfs_attr_set_args(
 	int			error, error2 = 0;
 
 	/*
-	 * If the attribute list is non-existent or a shortform list,
-	 * upgrade it to a single-leaf-block attribute list.
+	 * If the attribute list is non-existent or a shortform list, proceed to
+	 * upgrade it to a single-leaf-block attribute list.  Otherwise jump
+	 * straight to leaf handling
 	 */
-	if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL ||
+	if (!(dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL ||
 	    (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
-	     dp->i_d.di_anextents == 0)) {
+	     dp->i_d.di_anextents == 0)))
+		goto sf_to_leaf;
 
-		/*
-		 * Try to add the attr to the attribute list in the inode.
-		 */
-		error = xfs_attr_try_sf_addname(dp, args);
-		if (error != -ENOSPC) {
-			error2 = xfs_trans_commit(args->trans);
-			args->trans = NULL;
-			return error ? error : error2;
-		}
+	/*
+	 * Try to add the attr to the attribute list in the inode.
+	 */
+	error = xfs_attr_try_sf_addname(dp, args);
+	if (error != -ENOSPC) {
+		error2 = xfs_trans_commit(args->trans);
+		args->trans = NULL;
+		return error ? error : error2;
+	}
 
-		/*
-		 * It won't fit in the shortform, transform to a leaf block.
-		 * GROT: another possible req'mt for a double-split btree op.
-		 */
-		error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
-		if (error)
-			return error;
+	/*
+	 * It won't fit in the shortform, transform to a leaf block.
+	 * GROT: another possible req'mt for a double-split btree op.
+	 */
+	error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
+	if (error)
+		return error;
 
-		/*
-		 * Prevent the leaf buffer from being unlocked so that a
-		 * concurrent AIL push cannot grab the half-baked leaf
-		 * buffer and run into problems with the write verifier.
-		 * Once we're done rolling the transaction we can release
-		 * the hold and add the attr to the leaf.
-		 */
-		xfs_trans_bhold(args->trans, leaf_bp);
-		error = xfs_defer_finish(&args->trans);
-		xfs_trans_bhold_release(args->trans, leaf_bp);
-		if (error) {
-			xfs_trans_brelse(args->trans, leaf_bp);
-			return error;
-		}
+	/*
+	 * Prevent the leaf buffer from being unlocked so that a
+	 * concurrent AIL push cannot grab the half-baked leaf
+	 * buffer and run into problems with the write verifier.
+	 * Once we're done rolling the transaction we can release
+	 * the hold and add the attr to the leaf.
+	 */
+	xfs_trans_bhold(args->trans, leaf_bp);
+	error = xfs_defer_finish(&args->trans);
+	xfs_trans_bhold_release(args->trans, leaf_bp);
+	if (error) {
+		xfs_trans_brelse(args->trans, leaf_bp);
+		return error;
 	}
 
+sf_to_leaf:
+
 	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
 		error = xfs_attr_leaf_addname(args);
 		if (error != -ENOSPC)
-- 
2.7.4

[PATCH v6 15/16] xfs: Add delay ready attr remove routines

[Allison Collins]

This patch modifies the attr remove routines to be delay ready. This
means they no longer roll or commit transactions, but instead return
-EAGAIN to have the calling routine roll and refresh the transaction.
In this series, xfs_attr_remove_args has become xfs_attr_remove_iter,
which uses a sort of state machine like switch to keep track of where it
was when EAGAIN was returned. xfs_attr_node_removename has also been
modified to use the switch, and a  new version of xfs_attr_remove_args
consists of a simple loop to refresh the transaction until the operation
is completed.  A helper function xfs_attr_node_shrink has also been
added to help simplify xfs_attr_node_removename and reduce length.

This patch also adds a new struct xfs_delattr_context, which we will use
to keep track of the current state of an attribute operation. The new
xfs_delattr_state enum is used to track various operations that are in
progress so that we know not to repeat them, and resume where we left
off before EAGAIN was returned to cycle out the transaction. Other
members take the place of local variables that need to retain their
values across multiple function recalls.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c     | 181 +++++++++++++++++++++++++++++++++----------
 fs/xfs/libxfs/xfs_attr.h     |   1 +
 fs/xfs/libxfs/xfs_da_btree.h |  24 ++++++
 fs/xfs/scrub/common.c        |   2 +
 fs/xfs/xfs_acl.c             |   2 +
 fs/xfs/xfs_attr_list.c       |   1 +
 fs/xfs/xfs_ioctl.c           |   2 +
 fs/xfs/xfs_ioctl32.c         |   2 +
 fs/xfs/xfs_iops.c            |   2 +
 fs/xfs/xfs_xattr.c           |   1 +
 10 files changed, 176 insertions(+), 42 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index 90e0b2d..92929ad 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -371,11 +371,60 @@ xfs_has_attr(
  */
 int
 xfs_attr_remove_args(
+	struct xfs_da_args	*args)
+{
+	int			error = 0;
+	int			err2 = 0;
+
+	do {
+		error = xfs_attr_remove_iter(args);
+		if (error && error != -EAGAIN)
+			goto out;
+
+		if (args->dac.flags & XFS_DAC_FINISH_TRANS) {
+			args->dac.flags &= ~XFS_DAC_FINISH_TRANS;
+
+			err2 = xfs_defer_finish(&args->trans);
+			if (err2) {
+				error = err2;
+				goto out;
+			}
+		}
+
+		err2 = xfs_trans_roll_inode(&args->trans, args->dp);
+		if (err2) {
+			error = err2;
+			goto out;
+		}
+
+	} while (error == -EAGAIN);
+out:
+	return error;
+}
+
+/*
+ * Remove the attribute specified in @args.
+ * This routine is meant to function as a delayed operation, and may return
+ * -EGAIN when the transaction needs to be rolled.  Calling functions will need
+ * to handle this, and recall the function until a successful error code is
+ * returned.
+ */
+int
+xfs_attr_remove_iter(
 	struct xfs_da_args      *args)
 {
 	struct xfs_inode	*dp = args->dp;
 	int			error;
 
+	/* State machine switch */
+	switch (args->dac.dela_state) {
+	case XFS_DAS_RM_SHRINK:
+	case XFS_DAS_RM_NODE_BLKS:
+		goto node;
+	default:
+		break;
+	}
+
 	if (!xfs_inode_hasattr(dp)) {
 		error = -ENOATTR;
 	} else if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
@@ -384,6 +433,7 @@ xfs_attr_remove_args(
 	} else if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
 		error = xfs_attr_leaf_removename(args);
 	} else {
+node:
 		error = xfs_attr_node_removename(args);
 	}
 
@@ -887,9 +937,8 @@ xfs_attr_leaf_removename(
 		/* bp is gone due to xfs_da_shrink_inode */
 		if (error)
 			return error;
-		error = xfs_defer_finish(&args->trans);
-		if (error)
-			return error;
+
+		args->dac.flags |= XFS_DAC_FINISH_TRANS;
 	}
 	return 0;
 }
@@ -1233,6 +1282,42 @@ xfs_attr_init_unmapstate(
 	return 0;
 }
 
+/*
+ * Shrink an attribute from leaf to shortform
+ */
+STATIC int
+xfs_attr_node_shrink(
+	struct xfs_da_args	*args,
+	struct xfs_da_state     *state)
+{
+	struct xfs_inode	*dp = args->dp;
+	int			error, forkoff;
+	struct xfs_buf		*bp;
+
+	/*
+	 * Have to get rid of the copy of this dabuf in the state.
+	 */
+	ASSERT(state->path.active == 1);
+	ASSERT(state->path.blk[0].bp);
+	state->path.blk[0].bp = NULL;
+
+	error = xfs_attr3_leaf_read(args->trans, args->dp, 0, &bp);
+	if (error)
+		return error;
+
+	forkoff = xfs_attr_shortform_allfit(bp, dp);
+	if (forkoff) {
+		error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
+		/* bp is gone due to xfs_da_shrink_inode */
+		if (error)
+			return error;
+
+		args->dac.flags |= XFS_DAC_FINISH_TRANS;
+	} else
+		xfs_trans_brelse(args->trans, bp);
+
+	return 0;
+}
 
 /*
  * Remove a name from a B-tree attribute list.
@@ -1240,6 +1325,11 @@ xfs_attr_init_unmapstate(
  * This will involve walking down the Btree, and may involve joining
  * leaf nodes and even joining intermediate nodes up to and including
  * the root node (a special case of an intermediate node).
+ *
+ * This routine is meant to function as either an inline or delayed operation,
+ * and may return -EAGAIN when the transaction needs to be rolled.  Calling
+ * functions will need to handle this, and recall the function until a
+ * successful error code is returned.
  */
 STATIC int
 xfs_attr_node_removename(
@@ -1247,15 +1337,28 @@ xfs_attr_node_removename(
 {
 	struct xfs_da_state	*state;
 	struct xfs_da_state_blk	*blk;
-	struct xfs_buf		*bp;
-	int			retval, error, forkoff;
+	int			retval, error;
 	struct xfs_inode	*dp = args->dp;
 
 	trace_xfs_attr_node_removename(args);
+	state = args->dac.da_state;
+	blk = args->dac.blk;
+
+	/* State machine switch */
+	switch (args->dac.dela_state) {
+	case XFS_DAS_RM_NODE_BLKS:
+		goto rm_node_blks;
+	case XFS_DAS_RM_SHRINK:
+		goto rm_shrink;
+	default:
+		break;
+	}
 
 	error = xfs_attr_node_hasname(args, &state);
 	if (error != -EEXIST)
 		goto out;
+	else
+		error = 0;
 
 	/*
 	 * If there is an out-of-line value, de-allocate the blocks.
@@ -1265,18 +1368,36 @@ xfs_attr_node_removename(
 	blk = &state->path.blk[ state->path.active-1 ];
 	ASSERT(blk->bp != NULL);
 	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
+
+	/*
+	 * Store blk and state in the context incase we need to cycle out the
+	 * transaction
+	 */
+	args->dac.blk = blk;
+	args->dac.da_state = state;
+
 	if (args->rmtblkno > 0) {
 		error = xfs_attr_init_unmapstate(args, state);
 		if (error)
 			goto out;
 
-		error = xfs_trans_roll_inode(&args->trans, args->dp);
+		error = xfs_attr_rmtval_invalidate(args);
 		if (error)
 			goto out;
+	}
 
-		error = xfs_attr_rmtval_remove(args);
-		if (error)
-			goto out;
+rm_node_blks:
+
+	args->dac.dela_state = XFS_DAS_RM_NODE_BLKS;
+
+	if (args->rmtblkno > 0) {
+		error = xfs_attr_rmtval_unmap(args);
+
+		if (error) {
+			if (error == -EAGAIN)
+				args->dac.dela_state = XFS_DAS_RM_NODE_BLKS;
+			return error;
+		}
 
 		/*
 		 * Refill the state structure with buffers, the prior calls
@@ -1302,44 +1423,20 @@ xfs_attr_node_removename(
 		error = xfs_da3_join(state);
 		if (error)
 			goto out;
-		error = xfs_defer_finish(&args->trans);
-		if (error)
-			goto out;
-		/*
-		 * Commit the Btree join operation and start a new trans.
-		 */
-		error = xfs_trans_roll_inode(&args->trans, dp);
-		if (error)
-			goto out;
+
+		args->dac.flags |= XFS_DAC_FINISH_TRANS;
+		args->dac.dela_state = XFS_DAS_RM_SHRINK;
+		return -EAGAIN;
 	}
 
+rm_shrink:
+	args->dac.dela_state = XFS_DAS_RM_SHRINK;
+
 	/*
 	 * If the result is small enough, push it all into the inode.
 	 */
-	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
-		/*
-		 * Have to get rid of the copy of this dabuf in the state.
-		 */
-		ASSERT(state->path.active == 1);
-		ASSERT(state->path.blk[0].bp);
-		state->path.blk[0].bp = NULL;
-
-		error = xfs_attr3_leaf_read(args->trans, args->dp, 0, &bp);
-		if (error)
-			goto out;
-
-		if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
-			error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
-			/* bp is gone due to xfs_da_shrink_inode */
-			if (error)
-				goto out;
-			error = xfs_defer_finish(&args->trans);
-			if (error)
-				goto out;
-		} else
-			xfs_trans_brelse(args->trans, bp);
-	}
-	error = 0;
+	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK))
+		error = xfs_attr_node_shrink(args, state);
 
 out:
 	if (state)
diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
index ce7b039..ea873a5 100644
--- a/fs/xfs/libxfs/xfs_attr.h
+++ b/fs/xfs/libxfs/xfs_attr.h
@@ -155,6 +155,7 @@ int xfs_attr_set_args(struct xfs_da_args *args);
 int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
 int xfs_has_attr(struct xfs_da_args *args);
 int xfs_attr_remove_args(struct xfs_da_args *args);
+int xfs_attr_remove_iter(struct xfs_da_args *args);
 int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
 		  int flags, struct attrlist_cursor_kern *cursor);
 bool xfs_attr_namecheck(const void *name, size_t length);
diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
index 14f1be3..7fc87da 100644
--- a/fs/xfs/libxfs/xfs_da_btree.h
+++ b/fs/xfs/libxfs/xfs_da_btree.h
@@ -50,9 +50,33 @@ enum xfs_dacmp {
 };
 
 /*
+ * Enum values for xfs_delattr_context.da_state
+ */
+enum xfs_delattr_state {
+	XFS_DAS_RM_SHRINK	= 1, /* We are shrinking the tree */
+	XFS_DAS_RM_NODE_BLKS	= 2, /* We are removing node blocks */
+};
+
+/*
+ * Defines for xfs_delattr_context.flags
+ */
+#define	XFS_DAC_FINISH_TRANS	0x1 /* indicates to finish the transaction */
+
+/*
+ * Context used for keeping track of delayed attribute operations
+ */
+struct xfs_delattr_context {
+	struct xfs_da_state	*da_state;
+	struct xfs_da_state_blk *blk;
+	int			flags;
+	enum xfs_delattr_state	dela_state;
+};
+
+/*
  * Structure to ease passing around component names.
  */
 typedef struct xfs_da_args {
+	struct xfs_delattr_context dac; /* context used for delay attr ops */
 	struct xfs_da_geometry *geo;	/* da block geometry */
 	struct xfs_name	name;		/* name, length and argument  flags*/
 	uint8_t		filetype;	/* filetype of inode for directories */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 1887605..9a649d1 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -24,6 +24,8 @@
 #include "xfs_rmap_btree.h"
 #include "xfs_log.h"
 #include "xfs_trans_priv.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
 #include "xfs_attr.h"
 #include "xfs_reflink.h"
 #include "scrub/scrub.h"
diff --git a/fs/xfs/xfs_acl.c b/fs/xfs/xfs_acl.c
index 42ac847..d65e6d8 100644
--- a/fs/xfs/xfs_acl.c
+++ b/fs/xfs/xfs_acl.c
@@ -10,6 +10,8 @@
 #include "xfs_trans_resv.h"
 #include "xfs_mount.h"
 #include "xfs_inode.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
 #include "xfs_attr.h"
 #include "xfs_trace.h"
 #include "xfs_error.h"
diff --git a/fs/xfs/xfs_attr_list.c b/fs/xfs/xfs_attr_list.c
index d37743b..881b9a4 100644
--- a/fs/xfs/xfs_attr_list.c
+++ b/fs/xfs/xfs_attr_list.c
@@ -12,6 +12,7 @@
 #include "xfs_trans_resv.h"
 #include "xfs_mount.h"
 #include "xfs_da_format.h"
+#include "xfs_da_btree.h"
 #include "xfs_inode.h"
 #include "xfs_trans.h"
 #include "xfs_bmap.h"
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index 28c07c9..7c1d9da 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -15,6 +15,8 @@
 #include "xfs_iwalk.h"
 #include "xfs_itable.h"
 #include "xfs_error.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
 #include "xfs_attr.h"
 #include "xfs_bmap.h"
 #include "xfs_bmap_util.h"
diff --git a/fs/xfs/xfs_ioctl32.c b/fs/xfs/xfs_ioctl32.c
index 769581a..d504f8f 100644
--- a/fs/xfs/xfs_ioctl32.c
+++ b/fs/xfs/xfs_ioctl32.c
@@ -17,6 +17,8 @@
 #include "xfs_itable.h"
 #include "xfs_fsops.h"
 #include "xfs_rtalloc.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
 #include "xfs_attr.h"
 #include "xfs_ioctl.h"
 #include "xfs_ioctl32.h"
diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index e85bbf5..a2d299f 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -13,6 +13,8 @@
 #include "xfs_inode.h"
 #include "xfs_acl.h"
 #include "xfs_quota.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
 #include "xfs_attr.h"
 #include "xfs_trans.h"
 #include "xfs_trace.h"
diff --git a/fs/xfs/xfs_xattr.c b/fs/xfs/xfs_xattr.c
index 74133a5..d8dc72d 100644
--- a/fs/xfs/xfs_xattr.c
+++ b/fs/xfs/xfs_xattr.c
@@ -10,6 +10,7 @@
 #include "xfs_log_format.h"
 #include "xfs_da_format.h"
 #include "xfs_inode.h"
+#include "xfs_da_btree.h"
 #include "xfs_attr.h"
 #include "xfs_acl.h"
 
-- 
2.7.4

[PATCH v6 16/16] xfs: Add delay ready attr set routines

[Allison Collins]

This patch modifies the attr set routines to be delay ready. This means
they no longer roll or commit transactions, but instead return -EAGAIN
to have the calling routine roll and refresh the transaction.  In this
series, xfs_attr_set_args has become xfs_attr_set_iter, which uses a
state machine like switch to keep track of where it was when EAGAIN was
returned.

Part of xfs_attr_leaf_addname has been factored out into a new helper
function xfs_attr_leaf_try_add to allow transaction cycling between the
two routines.

Two new helper functions have been added: xfs_attr_rmtval_set_init and
xfs_attr_rmtval_set_blk.  They provide a subset of logic similar to
xfs_attr_rmtval_set, but they store the current block in the delay attr
context to allow the caller to roll the transaction between allocations.
This helps to simplify and consolidate code used by
xfs_attr_leaf_addname and xfs_attr_node_addname

Finally, xfs_attr_set_args has become a simple loop to refresh the
transaction until the operation is completed.

Signed-off-by: Allison Collins <allison.henderson@oracle.com>
---
 fs/xfs/libxfs/xfs_attr.c        | 347 +++++++++++++++++++++++++---------------
 fs/xfs/libxfs/xfs_attr.h        |   1 +
 fs/xfs/libxfs/xfs_attr_remote.c |  67 +++++++-
 fs/xfs/libxfs/xfs_attr_remote.h |   4 +
 fs/xfs/libxfs/xfs_da_btree.h    |  13 ++
 5 files changed, 301 insertions(+), 131 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index 92929ad..02a7f55 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -58,6 +58,7 @@ STATIC int xfs_attr_node_hasname(xfs_da_args_t *args,
 				 struct xfs_da_state **state);
 STATIC int xfs_attr_fillstate(xfs_da_state_t *state);
 STATIC int xfs_attr_refillstate(xfs_da_state_t *state);
+STATIC int xfs_attr_leaf_try_add(struct xfs_da_args *args, struct xfs_buf *bp);
 
 
 STATIC int
@@ -259,9 +260,86 @@ int
 xfs_attr_set_args(
 	struct xfs_da_args	*args)
 {
+	int			error = 0;
+	int			err2 = 0;
+	struct xfs_buf		*leaf_bp = NULL;
+
+	do {
+		error = xfs_attr_set_iter(args, &leaf_bp);
+		if (error && error != -EAGAIN)
+			goto out;
+
+		if (args->dac.flags & XFS_DAC_FINISH_TRANS) {
+			args->dac.flags &= ~XFS_DAC_FINISH_TRANS;
+
+			err2 = xfs_defer_finish(&args->trans);
+			if (err2) {
+				error = err2;
+				goto out;
+			}
+		}
+
+		err2 = xfs_trans_roll_inode(&args->trans, args->dp);
+		if (err2) {
+			error = err2;
+			goto out;
+		}
+
+		if (leaf_bp) {
+			xfs_trans_bjoin(args->trans, leaf_bp);
+			xfs_trans_bhold(args->trans, leaf_bp);
+		}
+
+	} while (error == -EAGAIN);
+
+out:
+	return error;
+}
+
+/*
+ * Set the attribute specified in @args.
+ * This routine is meant to function as a delayed operation, and may return
+ * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
+ * to handle this, and recall the function until a successful error code is
+ * returned.
+ */
+int
+xfs_attr_set_iter(
+	struct xfs_da_args	*args,
+	struct xfs_buf          **leaf_bp)
+{
 	struct xfs_inode	*dp = args->dp;
-	struct xfs_buf          *leaf_bp = NULL;
-	int			error, error2 = 0;
+	int			error = 0;
+	int			sf_size;
+
+	/* State machine switch */
+	switch (args->dac.dela_state) {
+	case XFS_DAS_SF_TO_LEAF:
+		goto sf_to_leaf;
+	case XFS_DAS_ALLOC_LEAF:
+	case XFS_DAS_FLIP_LFLAG:
+	case XFS_DAS_FOUND_LBLK:
+		goto leaf;
+	case XFS_DAS_FOUND_NBLK:
+	case XFS_DAS_FLIP_NFLAG:
+	case XFS_DAS_ALLOC_NODE:
+	case XFS_DAS_LEAF_TO_NODE:
+		goto node;
+	default:
+		break;
+	}
+
+	/*
+	 * New inodes may not have an attribute fork yet. So set the attribute
+	 * fork appropriately
+	 */
+	if (XFS_IFORK_Q((args->dp)) == 0) {
+		sf_size = sizeof(struct xfs_attr_sf_hdr) +
+		     XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
+		xfs_bmap_set_attrforkoff(args->dp, sf_size, NULL);
+		args->dp->i_afp = kmem_zone_zalloc(xfs_ifork_zone, 0);
+		args->dp->i_afp->if_flags = XFS_IFEXTENTS;
+	}
 
 	/*
 	 * If the attribute list is non-existent or a shortform list, proceed to
@@ -272,22 +350,20 @@ xfs_attr_set_args(
 	    (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
 	     dp->i_d.di_anextents == 0)))
 		goto sf_to_leaf;
-
 	/*
 	 * Try to add the attr to the attribute list in the inode.
 	 */
 	error = xfs_attr_try_sf_addname(dp, args);
-	if (error != -ENOSPC) {
-		error2 = xfs_trans_commit(args->trans);
-		args->trans = NULL;
-		return error ? error : error2;
-	}
+
+	/* Should only be 0, -EEXIST or ENOSPC */
+	if (error != -ENOSPC)
+		return error;
 
 	/*
 	 * It won't fit in the shortform, transform to a leaf block.
 	 * GROT: another possible req'mt for a double-split btree op.
 	 */
-	error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
+	error = xfs_attr_shortform_to_leaf(args, leaf_bp);
 	if (error)
 		return error;
 
@@ -295,42 +371,47 @@ xfs_attr_set_args(
 	 * Prevent the leaf buffer from being unlocked so that a
 	 * concurrent AIL push cannot grab the half-baked leaf
 	 * buffer and run into problems with the write verifier.
-	 * Once we're done rolling the transaction we can release
-	 * the hold and add the attr to the leaf.
 	 */
-	xfs_trans_bhold(args->trans, leaf_bp);
-	error = xfs_defer_finish(&args->trans);
-	xfs_trans_bhold_release(args->trans, leaf_bp);
-	if (error) {
-		xfs_trans_brelse(args->trans, leaf_bp);
-		return error;
-	}
+	xfs_trans_bhold(args->trans, *leaf_bp);
+	args->dac.flags |= XFS_DAC_FINISH_TRANS;
+	args->dac.dela_state = XFS_DAS_SF_TO_LEAF;
+	return -EAGAIN;
 
 sf_to_leaf:
 
-	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
-		error = xfs_attr_leaf_addname(args);
-		if (error != -ENOSPC)
-			return error;
-
-		/*
-		 * Commit that transaction so that the node_addname()
-		 * call can manage its own transactions.
-		 */
-		error = xfs_defer_finish(&args->trans);
-		if (error)
-			return error;
+	/*
+	 * After a shortform to leaf conversion, we need to hold the leaf and
+	 * cylce out the transaction.  When we get back, we need to release
+	 * the leaf.
+	 */
+	if (*leaf_bp != NULL) {
+		xfs_trans_brelse(args->trans, *leaf_bp);
+		*leaf_bp = NULL;
+	}
 
-		/*
-		 * Commit the current trans (including the inode) and
-		 * start a new one.
-		 */
-		error = xfs_trans_roll_inode(&args->trans, dp);
-		if (error)
+	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
+		error = xfs_attr_leaf_try_add(args, *leaf_bp);
+		switch (error) {
+		case -ENOSPC:
+			args->dac.flags |= XFS_DAC_FINISH_TRANS;
+			args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
+			return -EAGAIN;
+		case 0:
+			args->dac.dela_state = XFS_DAS_FOUND_LBLK;
+			return -EAGAIN;
+		default:
 			return error;
-
+		}
+leaf:
+		error = xfs_attr_leaf_addname(args);
+		if (error == -ENOSPC) {
+			args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
+			return -EAGAIN;
+		}
+		return error;
 	}
-
+	args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
+node:
 	error = xfs_attr_node_addname(args);
 	return error;
 }
@@ -759,27 +840,28 @@ xfs_attr_leaf_try_add(
  *
  * This leaf block cannot have a "remote" value, we only call this routine
  * if bmap_one_block() says there is only one block (ie: no remote blks).
+ *
+ * This routine is meant to function as a delayed operation, and may return
+ * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
+ * to handle this, and recall the function until a successful error code is
+ * returned.
  */
 STATIC int
 xfs_attr_leaf_addname(struct xfs_da_args	*args)
 {
-	int			error, forkoff;
 	struct xfs_buf		*bp = NULL;
+	int			error, forkoff;
 	struct xfs_inode	*dp = args->dp;
 
-	trace_xfs_attr_leaf_addname(args);
-
-	error = xfs_attr_leaf_try_add(args, bp);
-	if (error)
-		return error;
-
-	/*
-	 * Commit the transaction that added the attr name so that
-	 * later routines can manage their own transactions.
-	 */
-	error = xfs_trans_roll_inode(&args->trans, dp);
-	if (error)
-		return error;
+	/* State machine switch */
+	switch (args->dac.dela_state) {
+	case XFS_DAS_FLIP_LFLAG:
+		goto flip_flag;
+	case XFS_DAS_ALLOC_LEAF:
+		goto alloc_leaf;
+	default:
+		break;
+	}
 
 	/*
 	 * If there was an out-of-line value, allocate the blocks we
@@ -788,7 +870,28 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
 	 * maximum size of a transaction and/or hit a deadlock.
 	 */
 	if (args->rmtblkno > 0) {
-		error = xfs_attr_rmtval_set(args);
+
+		/* Open coded xfs_attr_rmtval_set without trans handling */
+		error = xfs_attr_rmtval_set_init(args);
+		if (error)
+			return error;
+
+		/*
+		 * Roll through the "value", allocating blocks on disk as
+		 * required.
+		 */
+alloc_leaf:
+		while (args->dac.blkcnt > 0) {
+			error = xfs_attr_rmtval_set_blk(args);
+			if (error)
+				return error;
+
+			args->dac.flags |= XFS_DAC_FINISH_TRANS;
+			args->dac.dela_state = XFS_DAS_ALLOC_LEAF;
+			return -EAGAIN;
+		}
+
+		error = xfs_attr_rmtval_set_value(args);
 		if (error)
 			return error;
 	}
@@ -807,13 +910,10 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
 		error = xfs_attr3_leaf_flipflags(args);
 		if (error)
 			return error;
-		/*
-		 * Commit the flag value change and start the next trans in
-		 * series.
-		 */
-		error = xfs_trans_roll_inode(&args->trans, args->dp);
-		if (error)
-			return error;
+
+		args->dac.dela_state = XFS_DAS_FLIP_LFLAG;
+		return -EAGAIN;
+flip_flag:
 
 		/*
 		 * Dismantle the "old" attribute/value pair by removing
@@ -844,34 +944,17 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
 		/*
 		 * If the result is small enough, shrink it all into the inode.
 		 */
-		if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
+		forkoff = xfs_attr_shortform_allfit(bp, dp);
+		if (forkoff)
 			error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
-			/* bp is gone due to xfs_da_shrink_inode */
-			if (error)
-				return error;
-			error = xfs_defer_finish(&args->trans);
-			if (error)
-				return error;
-		}
 
-		/*
-		 * Commit the remove and start the next trans in series.
-		 */
-		error = xfs_trans_roll_inode(&args->trans, dp);
+		args->dac.flags |= XFS_DAC_FINISH_TRANS;
 
 	} else if (args->rmtblkno > 0) {
 		/*
 		 * Added a "remote" value, just clear the incomplete flag.
 		 */
 		error = xfs_attr3_leaf_clearflag(args);
-		if (error)
-			return error;
-
-		/*
-		 * Commit the flag value change and start the next trans in
-		 * series.
-		 */
-		error = xfs_trans_roll_inode(&args->trans, args->dp);
 	}
 	return error;
 }
@@ -1020,16 +1103,22 @@ xfs_attr_node_hasname(
  *
  * "Remote" attribute values confuse the issue and atomic rename operations
  * add a whole extra layer of confusion on top of that.
+ *
+ * This routine is meant to function as a delayed operation, and may return
+ * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
+ * to handle this, and recall the function until a successful error code is
+ *returned.
  */
 STATIC int
 xfs_attr_node_addname(
 	struct xfs_da_args	*args)
 {
-	struct xfs_da_state	*state;
+	struct xfs_da_state	*state = NULL;
 	struct xfs_da_state_blk	*blk;
 	struct xfs_inode	*dp;
 	struct xfs_mount	*mp;
-	int			retval, error;
+	int			retval = 0;
+	int			error = 0;
 
 	trace_xfs_attr_node_addname(args);
 
@@ -1038,7 +1127,19 @@ xfs_attr_node_addname(
 	 */
 	dp = args->dp;
 	mp = dp->i_mount;
-restart:
+
+	/* State machine switch */
+	switch (args->dac.dela_state) {
+	case XFS_DAS_FLIP_NFLAG:
+		goto flip_flag;
+	case XFS_DAS_FOUND_NBLK:
+		goto found_nblk;
+	case XFS_DAS_ALLOC_NODE:
+		goto alloc_node;
+	default:
+		break;
+	}
+
 	/*
 	 * Search to see if name already exists, and get back a pointer
 	 * to where it should go.
@@ -1088,19 +1189,13 @@ xfs_attr_node_addname(
 			error = xfs_attr3_leaf_to_node(args);
 			if (error)
 				goto out;
-			error = xfs_defer_finish(&args->trans);
-			if (error)
-				goto out;
 
 			/*
-			 * Commit the node conversion and start the next
-			 * trans in the chain.
+			 * Restart routine from the top.  No need to set  the
+			 * state
 			 */
-			error = xfs_trans_roll_inode(&args->trans, dp);
-			if (error)
-				goto out;
-
-			goto restart;
+			args->dac.flags |= XFS_DAC_FINISH_TRANS;
+			return -EAGAIN;
 		}
 
 		/*
@@ -1112,9 +1207,7 @@ xfs_attr_node_addname(
 		error = xfs_da3_split(state);
 		if (error)
 			goto out;
-		error = xfs_defer_finish(&args->trans);
-		if (error)
-			goto out;
+		args->dac.flags |= XFS_DAC_FINISH_TRANS;
 	} else {
 		/*
 		 * Addition succeeded, update Btree hashvals.
@@ -1129,13 +1222,9 @@ xfs_attr_node_addname(
 	xfs_da_state_free(state);
 	state = NULL;
 
-	/*
-	 * Commit the leaf addition or btree split and start the next
-	 * trans in the chain.
-	 */
-	error = xfs_trans_roll_inode(&args->trans, dp);
-	if (error)
-		goto out;
+	args->dac.dela_state = XFS_DAS_FOUND_NBLK;
+	return -EAGAIN;
+found_nblk:
 
 	/*
 	 * If there was an out-of-line value, allocate the blocks we
@@ -1144,7 +1233,27 @@ xfs_attr_node_addname(
 	 * maximum size of a transaction and/or hit a deadlock.
 	 */
 	if (args->rmtblkno > 0) {
-		error = xfs_attr_rmtval_set(args);
+		/* Open coded xfs_attr_rmtval_set without trans handling */
+		error = xfs_attr_rmtval_set_init(args);
+		if (error)
+			return error;
+
+		/*
+		 * Roll through the "value", allocating blocks on disk as
+		 * required.
+		 */
+alloc_node:
+		while (args->dac.blkcnt > 0) {
+			error = xfs_attr_rmtval_set_blk(args);
+			if (error)
+				return error;
+
+			args->dac.flags |= XFS_DAC_FINISH_TRANS;
+			args->dac.dela_state = XFS_DAS_ALLOC_NODE;
+			return -EAGAIN;
+		}
+
+		error = xfs_attr_rmtval_set_value(args);
 		if (error)
 			return error;
 	}
@@ -1167,10 +1276,9 @@ xfs_attr_node_addname(
 		 * Commit the flag value change and start the next trans in
 		 * series
 		 */
-		error = xfs_trans_roll_inode(&args->trans, args->dp);
-		if (error)
-			goto out;
-
+		args->dac.dela_state = XFS_DAS_FLIP_NFLAG;
+		return -EAGAIN;
+flip_flag:
 		/*
 		 * Dismantle the "old" attribute/value pair by removing
 		 * a "remote" value (if it exists).
@@ -1199,7 +1307,6 @@ xfs_attr_node_addname(
 		error = xfs_da3_node_lookup_int(state, &retval);
 		if (error)
 			goto out;
-
 		/*
 		 * Remove the name and update the hashvals in the tree.
 		 */
@@ -1207,7 +1314,6 @@ xfs_attr_node_addname(
 		ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
 		error = xfs_attr3_leaf_remove(blk->bp, args);
 		xfs_da3_fixhashpath(state, &state->path);
-
 		/*
 		 * Check to see if the tree needs to be collapsed.
 		 */
@@ -1215,18 +1321,9 @@ xfs_attr_node_addname(
 			error = xfs_da3_join(state);
 			if (error)
 				goto out;
-			error = xfs_defer_finish(&args->trans);
-			if (error)
-				goto out;
-		}
-
-		/*
-		 * Commit and start the next trans in the chain.
-		 */
-		error = xfs_trans_roll_inode(&args->trans, dp);
-		if (error)
-			goto out;
 
+			args->dac.flags |= XFS_DAC_FINISH_TRANS;
+		}
 	} else if (args->rmtblkno > 0) {
 		/*
 		 * Added a "remote" value, just clear the incomplete flag.
@@ -1234,14 +1331,6 @@ xfs_attr_node_addname(
 		error = xfs_attr3_leaf_clearflag(args);
 		if (error)
 			goto out;
-
-		 /*
-		  * Commit the flag value change and start the next trans in
-		  * series.
-		  */
-		error = xfs_trans_roll_inode(&args->trans, args->dp);
-		if (error)
-			goto out;
 	}
 	retval = error = 0;
 
diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
index ea873a5..f450d8c 100644
--- a/fs/xfs/libxfs/xfs_attr.h
+++ b/fs/xfs/libxfs/xfs_attr.h
@@ -152,6 +152,7 @@ int xfs_attr_get(struct xfs_inode *ip, struct xfs_name *name,
 int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
 		 unsigned char *value, int valuelen, int flags);
 int xfs_attr_set_args(struct xfs_da_args *args);
+int xfs_attr_set_iter(struct xfs_da_args *args, struct xfs_buf **leaf_bp);
 int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
 int xfs_has_attr(struct xfs_da_args *args);
 int xfs_attr_remove_args(struct xfs_da_args *args);
diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
index f2ee0b8..ad4f14a 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.c
+++ b/fs/xfs/libxfs/xfs_attr_remote.c
@@ -428,7 +428,7 @@ xfs_attr_rmtval_get(
  * Find a "hole" in the attribute address space large enough for us to drop the
  * new attribute's value into
  */
-STATIC int
+int
 xfs_attr_rmt_find_hole(
 	struct xfs_da_args	*args)
 {
@@ -455,7 +455,7 @@ xfs_attr_rmt_find_hole(
 	return 0;
 }
 
-STATIC int
+int
 xfs_attr_rmtval_set_value(
 	struct xfs_da_args	*args)
 {
@@ -588,6 +588,69 @@ xfs_attr_rmtval_set(
 }
 
 /*
+ * Find a hole for the attr and store it in the delayed attr context.  This
+ * initializes the context to roll through allocating an attr extent for a
+ * delayed attr operation
+ */
+int
+xfs_attr_rmtval_set_init(
+	struct xfs_da_args	*args)
+{
+	struct xfs_bmbt_irec	*map = &args->dac.map;
+	int error;
+
+	args->dac.lblkno = 0;
+	args->dac.lfileoff = 0;
+	args->dac.blkcnt = 0;
+	args->rmtblkcnt = 0;
+	args->rmtblkno = 0;
+	memset(map, 0, sizeof(struct xfs_bmbt_irec));
+
+	error = xfs_attr_rmt_find_hole(args);
+	if (error)
+		return error;
+
+	args->dac.blkcnt = args->rmtblkcnt;
+	args->dac.lblkno = args->rmtblkno;
+
+	return error;
+}
+
+/*
+ * Write one block of the value associated with an attribute into the
+ * out-of-line buffer that we have defined for it. This is similar to a subset
+ * of xfs_attr_rmtval_set, but records the current block to the delayed attr
+ * context, and leaves transaction handling to the caller.
+ */
+int
+xfs_attr_rmtval_set_blk(
+	struct xfs_da_args	*args)
+{
+	struct xfs_inode	*dp = args->dp;
+	struct xfs_bmbt_irec	*map = &args->dac.map;
+	int nmap;
+	int error;
+
+	nmap = 1;
+	error = xfs_bmapi_write(args->trans, dp,
+		  (xfs_fileoff_t)args->dac.lblkno,
+		  args->dac.blkcnt, XFS_BMAPI_ATTRFORK,
+		  args->total, map, &nmap);
+	if (error)
+		return error;
+
+	ASSERT(nmap == 1);
+	ASSERT((map->br_startblock != DELAYSTARTBLOCK) &&
+	       (map->br_startblock != HOLESTARTBLOCK));
+
+	/* roll attribute extent map forwards */
+	args->dac.lblkno += map->br_blockcount;
+	args->dac.blkcnt -= map->br_blockcount;
+
+	return 0;
+}
+
+/*
  * Remove the value associated with an attribute by deleting the
  * out-of-line buffer that it is stored on.
  */
diff --git a/fs/xfs/libxfs/xfs_attr_remote.h b/fs/xfs/libxfs/xfs_attr_remote.h
index 7ab3770..ab03519 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.h
+++ b/fs/xfs/libxfs/xfs_attr_remote.h
@@ -13,4 +13,8 @@ int xfs_attr_rmtval_set(struct xfs_da_args *args);
 int xfs_attr_rmtval_remove(struct xfs_da_args *args);
 int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
 int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
+int xfs_attr_rmt_find_hole(struct xfs_da_args *args);
+int xfs_attr_rmtval_set_value(struct xfs_da_args *args);
+int xfs_attr_rmtval_set_blk(struct xfs_da_args *args);
+int xfs_attr_rmtval_set_init(struct xfs_da_args *args);
 #endif /* __XFS_ATTR_REMOTE_H__ */
diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
index 7fc87da..9943062 100644
--- a/fs/xfs/libxfs/xfs_da_btree.h
+++ b/fs/xfs/libxfs/xfs_da_btree.h
@@ -55,6 +55,14 @@ enum xfs_dacmp {
 enum xfs_delattr_state {
 	XFS_DAS_RM_SHRINK	= 1, /* We are shrinking the tree */
 	XFS_DAS_RM_NODE_BLKS	= 2, /* We are removing node blocks */
+	XFS_DAS_SF_TO_LEAF	= 3, /* Converted short form to leaf */
+	XFS_DAS_FOUND_LBLK	= 4, /* We found leaf blk for attr */
+	XFS_DAS_LEAF_TO_NODE	= 5, /* Converted leaf to node */
+	XFS_DAS_FOUND_NBLK	= 6, /* We found node blk for attr */
+	XFS_DAS_ALLOC_LEAF	= 7, /* We are allocating leaf blocks */
+	XFS_DAS_FLIP_LFLAG	= 8, /* Flipped leaf INCOMPLETE attr flag */
+	XFS_DAS_ALLOC_NODE	= 9, /* We are allocating node blocks */
+	XFS_DAS_FLIP_NFLAG	= 10,/* Flipped node INCOMPLETE attr flag */
 };
 
 /*
@@ -66,8 +74,13 @@ enum xfs_delattr_state {
  * Context used for keeping track of delayed attribute operations
  */
 struct xfs_delattr_context {
+	struct xfs_bmbt_irec	map;
+	struct xfs_buf		*leaf_bp;
+	xfs_fileoff_t		lfileoff;
 	struct xfs_da_state	*da_state;
 	struct xfs_da_state_blk *blk;
+	xfs_dablk_t		lblkno;
+	int			blkcnt;
 	int			flags;
 	enum xfs_delattr_state	dela_state;
 };
-- 
2.7.4

Re: [PATCH v6 01/16] xfs: Replace attribute parameters with struct xfs_name

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:20PM -0700, Allison Collins wrote:
> This patch replaces the attribute name and length parameters with a
> single struct xfs_name parameter.  This helps to clean up the numbers of
> parameters being passed around and pre-simplifies the code some.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c  | 22 +++++++++-------------
>  fs/xfs/libxfs/xfs_attr.h  | 12 +++++-------
>  fs/xfs/libxfs/xfs_types.c | 11 +++++++++++
>  fs/xfs/libxfs/xfs_types.h |  1 +
>  fs/xfs/xfs_acl.c          | 25 +++++++++++--------------
>  fs/xfs/xfs_ioctl.c        | 23 +++++++++++++----------
>  fs/xfs/xfs_iops.c         |  6 +++---
>  fs/xfs/xfs_xattr.c        | 28 ++++++++++++++++------------
>  8 files changed, 69 insertions(+), 59 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index e614972..6717f47 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -61,8 +61,7 @@ STATIC int
>  xfs_attr_args_init(
>  	struct xfs_da_args	*args,
>  	struct xfs_inode	*dp,
> -	const unsigned char	*name,
> -	size_t			namelen,
> +	struct xfs_name		*name,
>  	int			flags)
>  {
>  
> @@ -74,8 +73,8 @@ xfs_attr_args_init(
>  	args->whichfork = XFS_ATTR_FORK;
>  	args->dp = dp;
>  	args->flags = flags;
> -	args->name = name;
> -	args->namelen = namelen;
> +	args->name = name->name;
> +	args->namelen = name->len;
>  	if (args->namelen >= MAXNAMELEN)
>  		return -EFAULT;		/* match IRIX behaviour */
>  
> @@ -139,8 +138,7 @@ xfs_attr_get_ilocked(
>  int
>  xfs_attr_get(
>  	struct xfs_inode	*ip,
> -	const unsigned char	*name,
> -	size_t			namelen,
> +	struct xfs_name		*name,
>  	unsigned char		**value,
>  	int			*valuelenp,
>  	int			flags)
> @@ -156,7 +154,7 @@ xfs_attr_get(
>  	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
>  		return -EIO;
>  
> -	error = xfs_attr_args_init(&args, ip, name, namelen, flags);
> +	error = xfs_attr_args_init(&args, ip, name, flags);
>  	if (error)
>  		return error;
>  
> @@ -339,8 +337,7 @@ xfs_attr_remove_args(
>  int
>  xfs_attr_set(
>  	struct xfs_inode	*dp,
> -	const unsigned char	*name,
> -	size_t			namelen,
> +	struct xfs_name		*name,
>  	unsigned char		*value,
>  	int			valuelen,
>  	int			flags)
> @@ -356,7 +353,7 @@ xfs_attr_set(
>  	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
>  		return -EIO;
>  
> -	error = xfs_attr_args_init(&args, dp, name, namelen, flags);
> +	error = xfs_attr_args_init(&args, dp, name, flags);
>  	if (error)
>  		return error;
>  
> @@ -444,8 +441,7 @@ xfs_attr_set(
>  int
>  xfs_attr_remove(
>  	struct xfs_inode	*dp,
> -	const unsigned char	*name,
> -	size_t			namelen,
> +	struct xfs_name		*name,
>  	int			flags)
>  {
>  	struct xfs_mount	*mp = dp->i_mount;
> @@ -457,7 +453,7 @@ xfs_attr_remove(
>  	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
>  		return -EIO;
>  
> -	error = xfs_attr_args_init(&args, dp, name, namelen, flags);
> +	error = xfs_attr_args_init(&args, dp, name, flags);
>  	if (error)
>  		return error;
>  
> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
> index 4243b22..35043db 100644
> --- a/fs/xfs/libxfs/xfs_attr.h
> +++ b/fs/xfs/libxfs/xfs_attr.h
> @@ -147,14 +147,12 @@ int xfs_attr_list_int_ilocked(struct xfs_attr_list_context *);
>  int xfs_attr_list_int(struct xfs_attr_list_context *);
>  int xfs_inode_hasattr(struct xfs_inode *ip);
>  int xfs_attr_get_ilocked(struct xfs_inode *ip, struct xfs_da_args *args);
> -int xfs_attr_get(struct xfs_inode *ip, const unsigned char *name,
> -		 size_t namelen, unsigned char **value, int *valuelenp,
> -		 int flags);
> -int xfs_attr_set(struct xfs_inode *dp, const unsigned char *name,
> -		 size_t namelen, unsigned char *value, int valuelen, int flags);
> +int xfs_attr_get(struct xfs_inode *ip, struct xfs_name *name,
> +		 unsigned char **value, int *valuelenp, int flags);
> +int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
> +		 unsigned char *value, int valuelen, int flags);
>  int xfs_attr_set_args(struct xfs_da_args *args);
> -int xfs_attr_remove(struct xfs_inode *dp, const unsigned char *name,
> -		    size_t namelen, int flags);
> +int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
>  int xfs_attr_remove_args(struct xfs_da_args *args);
>  int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
>  		  int flags, struct attrlist_cursor_kern *cursor);
> diff --git a/fs/xfs/libxfs/xfs_types.c b/fs/xfs/libxfs/xfs_types.c
> index 4f59554..781a5a9 100644
> --- a/fs/xfs/libxfs/xfs_types.c
> +++ b/fs/xfs/libxfs/xfs_types.c
> @@ -12,6 +12,17 @@
>  #include "xfs_bit.h"
>  #include "xfs_mount.h"
>  
> +/* Initialize a struct xfs_name with a null terminated string name */
> +void
> +xfs_name_init(
> +	struct xfs_name *xname,
> +	const char *name)

Indent the variable names, please.

Though I have this feeling that the huge series from hch that I just
reviewed will make this patch unnecessary...

--D

> +{
> +	xname->name = (unsigned char *)name;
> +	xname->len = strlen(name);
> +	xname->type = 0;
> +}
> +
>  /* Find the size of the AG, in blocks. */
>  xfs_agblock_t
>  xfs_ag_block_count(
> diff --git a/fs/xfs/libxfs/xfs_types.h b/fs/xfs/libxfs/xfs_types.h
> index 397d947..b94acb5 100644
> --- a/fs/xfs/libxfs/xfs_types.h
> +++ b/fs/xfs/libxfs/xfs_types.h
> @@ -180,6 +180,7 @@ enum xfs_ag_resv_type {
>   */
>  struct xfs_mount;
>  
> +void xfs_name_init(struct xfs_name *xname, const char *name);
>  xfs_agblock_t xfs_ag_block_count(struct xfs_mount *mp, xfs_agnumber_t agno);
>  bool xfs_verify_agbno(struct xfs_mount *mp, xfs_agnumber_t agno,
>  		xfs_agblock_t agbno);
> diff --git a/fs/xfs/xfs_acl.c b/fs/xfs/xfs_acl.c
> index cd743fad..42ac847 100644
> --- a/fs/xfs/xfs_acl.c
> +++ b/fs/xfs/xfs_acl.c
> @@ -123,7 +123,7 @@ xfs_get_acl(struct inode *inode, int type)
>  	struct xfs_inode *ip = XFS_I(inode);
>  	struct posix_acl *acl = NULL;
>  	struct xfs_acl *xfs_acl = NULL;
> -	unsigned char *ea_name;
> +	struct xfs_name name;
>  	int error;
>  	int len;
>  
> @@ -131,10 +131,10 @@ xfs_get_acl(struct inode *inode, int type)
>  
>  	switch (type) {
>  	case ACL_TYPE_ACCESS:
> -		ea_name = SGI_ACL_FILE;
> +		xfs_name_init(&name, SGI_ACL_FILE);
>  		break;
>  	case ACL_TYPE_DEFAULT:
> -		ea_name = SGI_ACL_DEFAULT;
> +		xfs_name_init(&name, SGI_ACL_DEFAULT);
>  		break;
>  	default:
>  		BUG();
> @@ -145,9 +145,8 @@ xfs_get_acl(struct inode *inode, int type)
>  	 * go out to the disk.
>  	 */
>  	len = XFS_ACL_MAX_SIZE(ip->i_mount);
> -	error = xfs_attr_get(ip, ea_name, strlen(ea_name),
> -				(unsigned char **)&xfs_acl, &len,
> -				ATTR_ALLOC | ATTR_ROOT);
> +	error = xfs_attr_get(ip, &name, (unsigned char **)&xfs_acl, &len,
> +			     ATTR_ALLOC | ATTR_ROOT);
>  	if (error) {
>  		/*
>  		 * If the attribute doesn't exist make sure we have a negative
> @@ -167,17 +166,17 @@ int
>  __xfs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
>  {
>  	struct xfs_inode *ip = XFS_I(inode);
> -	unsigned char *ea_name;
> +	struct xfs_name name;
>  	int error;
>  
>  	switch (type) {
>  	case ACL_TYPE_ACCESS:
> -		ea_name = SGI_ACL_FILE;
> +		xfs_name_init(&name, SGI_ACL_FILE);
>  		break;
>  	case ACL_TYPE_DEFAULT:
>  		if (!S_ISDIR(inode->i_mode))
>  			return acl ? -EACCES : 0;
> -		ea_name = SGI_ACL_DEFAULT;
> +		xfs_name_init(&name, SGI_ACL_DEFAULT);
>  		break;
>  	default:
>  		return -EINVAL;
> @@ -197,17 +196,15 @@ __xfs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
>  		len -= sizeof(struct xfs_acl_entry) *
>  			 (XFS_ACL_MAX_ENTRIES(ip->i_mount) - acl->a_count);
>  
> -		error = xfs_attr_set(ip, ea_name, strlen(ea_name),
> -				     (unsigned char *)xfs_acl, len, ATTR_ROOT);
> +		error = xfs_attr_set(ip, &name, (unsigned char *)xfs_acl, len,
> +				     ATTR_ROOT);
>  
>  		kmem_free(xfs_acl);
>  	} else {
>  		/*
>  		 * A NULL ACL argument means we want to remove the ACL.
>  		 */
> -		error = xfs_attr_remove(ip, ea_name,
> -					strlen(ea_name),
> -					ATTR_ROOT);
> +		error = xfs_attr_remove(ip, &name, ATTR_ROOT);
>  
>  		/*
>  		 * If the attribute didn't exist to start with that's fine.
> diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> index d42de92..28c07c9 100644
> --- a/fs/xfs/xfs_ioctl.c
> +++ b/fs/xfs/xfs_ioctl.c
> @@ -357,7 +357,9 @@ xfs_attrmulti_attr_get(
>  {
>  	unsigned char		*kbuf;
>  	int			error = -EFAULT;
> -	size_t			namelen;
> +	struct xfs_name		xname;
> +
> +	xfs_name_init(&xname, name);
>  
>  	if (*len > XFS_XATTR_SIZE_MAX)
>  		return -EINVAL;
> @@ -365,9 +367,7 @@ xfs_attrmulti_attr_get(
>  	if (!kbuf)
>  		return -ENOMEM;
>  
> -	namelen = strlen(name);
> -	error = xfs_attr_get(XFS_I(inode), name, namelen, &kbuf, (int *)len,
> -			     flags);
> +	error = xfs_attr_get(XFS_I(inode), &xname, &kbuf, (int *)len, flags);
>  	if (error)
>  		goto out_kfree;
>  
> @@ -389,7 +389,9 @@ xfs_attrmulti_attr_set(
>  {
>  	unsigned char		*kbuf;
>  	int			error;
> -	size_t			namelen;
> +	struct xfs_name		xname;
> +
> +	xfs_name_init(&xname, name);
>  
>  	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
>  		return -EPERM;
> @@ -400,8 +402,7 @@ xfs_attrmulti_attr_set(
>  	if (IS_ERR(kbuf))
>  		return PTR_ERR(kbuf);
>  
> -	namelen = strlen(name);
> -	error = xfs_attr_set(XFS_I(inode), name, namelen, kbuf, len, flags);
> +	error = xfs_attr_set(XFS_I(inode), &xname, kbuf, len, flags);
>  	if (!error)
>  		xfs_forget_acl(inode, name, flags);
>  	kfree(kbuf);
> @@ -415,12 +416,14 @@ xfs_attrmulti_attr_remove(
>  	uint32_t		flags)
>  {
>  	int			error;
> -	size_t			namelen;
> +	struct xfs_name		xname;
> +
> +	xfs_name_init(&xname, name);
>  
>  	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
>  		return -EPERM;
> -	namelen = strlen(name);
> -	error = xfs_attr_remove(XFS_I(inode), name, namelen, flags);
> +
> +	error = xfs_attr_remove(XFS_I(inode), &xname, flags);
>  	if (!error)
>  		xfs_forget_acl(inode, name, flags);
>  	return error;
> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> index 81f2f93..e85bbf5 100644
> --- a/fs/xfs/xfs_iops.c
> +++ b/fs/xfs/xfs_iops.c
> @@ -48,11 +48,11 @@ xfs_initxattrs(
>  	const struct xattr	*xattr;
>  	struct xfs_inode	*ip = XFS_I(inode);
>  	int			error = 0;
> +	struct xfs_name		name;
>  
>  	for (xattr = xattr_array; xattr->name != NULL; xattr++) {
> -		error = xfs_attr_set(ip, xattr->name,
> -				     strlen(xattr->name),
> -				     xattr->value, xattr->value_len,
> +		xfs_name_init(&name, xattr->name);
> +		error = xfs_attr_set(ip, &name, xattr->value, xattr->value_len,
>  				     ATTR_SECURE);
>  		if (error < 0)
>  			break;
> diff --git a/fs/xfs/xfs_xattr.c b/fs/xfs/xfs_xattr.c
> index b0fedb5..74133a5 100644
> --- a/fs/xfs/xfs_xattr.c
> +++ b/fs/xfs/xfs_xattr.c
> @@ -21,10 +21,12 @@ static int
>  xfs_xattr_get(const struct xattr_handler *handler, struct dentry *unused,
>  		struct inode *inode, const char *name, void *value, size_t size)
>  {
> -	int xflags = handler->flags;
> -	struct xfs_inode *ip = XFS_I(inode);
> -	int error, asize = size;
> -	size_t namelen = strlen(name);
> +	int			xflags = handler->flags;
> +	struct xfs_inode	*ip = XFS_I(inode);
> +	int			error, asize = size;
> +	struct xfs_name		xname;
> +
> +	xfs_name_init(&xname, name);
>  
>  	/* Convert Linux syscall to XFS internal ATTR flags */
>  	if (!size) {
> @@ -32,8 +34,8 @@ xfs_xattr_get(const struct xattr_handler *handler, struct dentry *unused,
>  		value = NULL;
>  	}
>  
> -	error = xfs_attr_get(ip, name, namelen, (unsigned char **)&value,
> -			     &asize, xflags);
> +	error = xfs_attr_get(ip, &xname, (unsigned char **)&value, &asize,
> +			     xflags);
>  	if (error)
>  		return error;
>  	return asize;
> @@ -69,7 +71,9 @@ xfs_xattr_set(const struct xattr_handler *handler, struct dentry *unused,
>  	int			xflags = handler->flags;
>  	struct xfs_inode	*ip = XFS_I(inode);
>  	int			error;
> -	size_t			namelen = strlen(name);
> +	struct xfs_name		xname;
> +
> +	xfs_name_init(&xname, name);
>  
>  	/* Convert Linux syscall to XFS internal ATTR flags */
>  	if (flags & XATTR_CREATE)
> @@ -77,11 +81,11 @@ xfs_xattr_set(const struct xattr_handler *handler, struct dentry *unused,
>  	if (flags & XATTR_REPLACE)
>  		xflags |= ATTR_REPLACE;
>  
> -	if (value)
> -		error = xfs_attr_set(ip, name, namelen, (void *)value, size,
> -				xflags);
> -	else
> -		error = xfs_attr_remove(ip, name, namelen, xflags);
> +        if (value)
> +               error = xfs_attr_set(ip, &xname, (void *)value, size, xflags);
> +        else
> +               error = xfs_attr_remove(ip, &xname, xflags);
> +
>  	if (!error)
>  		xfs_forget_acl(inode, name, xflags);
>  
> -- 
> 2.7.4
> 

Re: [PATCH v6 02/16] xfs: Embed struct xfs_name in xfs_da_args

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:21PM -0700, Allison Collins wrote:
> This patch embeds an xfs_name in xfs_da_args, replacing the name,
> namelen, and flags members.  This helps to clean up the xfs_da_args
> structure and make it more uniform with the new xfs_name parameter being
> passed around.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> Reviewed-by: Brian Foster <bfoster@redhat.com>

Looks ok,
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>

(I'm not really sure /what/ tag to apply to this -- it looks reasonably
sane, it conflicts with a separate proposed cleanup series that would
make some of this rather less awkward, and yet at the same time I've not
actually put hch's series in for-next pending questions...)

--D

> ---
>  fs/xfs/libxfs/xfs_attr.c        |  37 +++++++-------
>  fs/xfs/libxfs/xfs_attr_leaf.c   | 104 +++++++++++++++++++++-------------------
>  fs/xfs/libxfs/xfs_attr_remote.c |   2 +-
>  fs/xfs/libxfs/xfs_da_btree.c    |   6 ++-
>  fs/xfs/libxfs/xfs_da_btree.h    |   4 +-
>  fs/xfs/libxfs/xfs_dir2.c        |  18 +++----
>  fs/xfs/libxfs/xfs_dir2_block.c  |   6 +--
>  fs/xfs/libxfs/xfs_dir2_leaf.c   |   6 +--
>  fs/xfs/libxfs/xfs_dir2_node.c   |   8 ++--
>  fs/xfs/libxfs/xfs_dir2_sf.c     |  30 ++++++------
>  fs/xfs/scrub/attr.c             |  12 ++---
>  fs/xfs/xfs_trace.h              |  20 ++++----
>  12 files changed, 130 insertions(+), 123 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index 6717f47..9acdb23 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -72,13 +72,12 @@ xfs_attr_args_init(
>  	args->geo = dp->i_mount->m_attr_geo;
>  	args->whichfork = XFS_ATTR_FORK;
>  	args->dp = dp;
> -	args->flags = flags;
> -	args->name = name->name;
> -	args->namelen = name->len;
> -	if (args->namelen >= MAXNAMELEN)
> +	memcpy(&args->name, name, sizeof(struct xfs_name));
> +	args->name.type = flags;
> +	if (args->name.len >= MAXNAMELEN)
>  		return -EFAULT;		/* match IRIX behaviour */
>  
> -	args->hashval = xfs_da_hashname(args->name, args->namelen);
> +	args->hashval = xfs_da_hashname(args->name.name, args->name.len);
>  	return 0;
>  }
>  
> @@ -236,7 +235,7 @@ xfs_attr_try_sf_addname(
>  	 * Commit the shortform mods, and we're done.
>  	 * NOTE: this is also the error path (EEXIST, etc).
>  	 */
> -	if (!error && (args->flags & ATTR_KERNOTIME) == 0)
> +	if (!error && (args->name.type & ATTR_KERNOTIME) == 0)
>  		xfs_trans_ichgtime(args->trans, dp, XFS_ICHGTIME_CHG);
>  
>  	if (mp->m_flags & XFS_MOUNT_WSYNC)
> @@ -357,6 +356,9 @@ xfs_attr_set(
>  	if (error)
>  		return error;
>  
> +	/* Use name now stored in args */
> +	name = &args.name;
> +
>  	args.value = value;
>  	args.valuelen = valuelen;
>  	args.op_flags = XFS_DA_OP_ADDNAME | XFS_DA_OP_OKNOENT;
> @@ -372,7 +374,7 @@ xfs_attr_set(
>  	 */
>  	if (XFS_IFORK_Q(dp) == 0) {
>  		int sf_size = sizeof(xfs_attr_sf_hdr_t) +
> -			XFS_ATTR_SF_ENTSIZE_BYNAME(args.namelen, valuelen);
> +			XFS_ATTR_SF_ENTSIZE_BYNAME(name->len, valuelen);
>  
>  		error = xfs_bmap_add_attrfork(dp, sf_size, rsvd);
>  		if (error)
> @@ -457,6 +459,9 @@ xfs_attr_remove(
>  	if (error)
>  		return error;
>  
> +	/* Use name now stored in args */
> +	name = &args.name;
> +
>  	/*
>  	 * we have no control over the attribute names that userspace passes us
>  	 * to remove, so we have to allow the name lookup prior to attribute
> @@ -532,10 +537,10 @@ xfs_attr_shortform_addname(xfs_da_args_t *args)
>  	trace_xfs_attr_sf_addname(args);
>  
>  	retval = xfs_attr_shortform_lookup(args);
> -	if ((args->flags & ATTR_REPLACE) && (retval == -ENOATTR)) {
> +	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
>  		return retval;
>  	} else if (retval == -EEXIST) {
> -		if (args->flags & ATTR_CREATE)
> +		if (args->name.type & ATTR_CREATE)
>  			return retval;
>  		retval = xfs_attr_shortform_remove(args);
>  		if (retval)
> @@ -545,15 +550,15 @@ xfs_attr_shortform_addname(xfs_da_args_t *args)
>  		 * that the leaf format add routine won't trip over the attr
>  		 * not being around.
>  		 */
> -		args->flags &= ~ATTR_REPLACE;
> +		args->name.type &= ~ATTR_REPLACE;
>  	}
>  
> -	if (args->namelen >= XFS_ATTR_SF_ENTSIZE_MAX ||
> +	if (args->name.len >= XFS_ATTR_SF_ENTSIZE_MAX ||
>  	    args->valuelen >= XFS_ATTR_SF_ENTSIZE_MAX)
>  		return -ENOSPC;
>  
>  	newsize = XFS_ATTR_SF_TOTSIZE(args->dp);
> -	newsize += XFS_ATTR_SF_ENTSIZE_BYNAME(args->namelen, args->valuelen);
> +	newsize += XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
>  
>  	forkoff = xfs_attr_shortform_bytesfit(args->dp, newsize);
>  	if (!forkoff)
> @@ -598,11 +603,11 @@ xfs_attr_leaf_addname(
>  	 * the given flags produce an error or call for an atomic rename.
>  	 */
>  	retval = xfs_attr3_leaf_lookup_int(bp, args);
> -	if ((args->flags & ATTR_REPLACE) && (retval == -ENOATTR)) {
> +	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
>  		xfs_trans_brelse(args->trans, bp);
>  		return retval;
>  	} else if (retval == -EEXIST) {
> -		if (args->flags & ATTR_CREATE) {	/* pure create op */
> +		if (args->name.type & ATTR_CREATE) {	/* pure create op */
>  			xfs_trans_brelse(args->trans, bp);
>  			return retval;
>  		}
> @@ -872,10 +877,10 @@ xfs_attr_node_addname(
>  		goto out;
>  	blk = &state->path.blk[ state->path.active-1 ];
>  	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
> -	if ((args->flags & ATTR_REPLACE) && (retval == -ENOATTR)) {
> +	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
>  		goto out;
>  	} else if (retval == -EEXIST) {
> -		if (args->flags & ATTR_CREATE)
> +		if (args->name.type & ATTR_CREATE)
>  			goto out;
>  
>  		trace_xfs_attr_node_replace(args);
> diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
> index fed537a..cb5ef66 100644
> --- a/fs/xfs/libxfs/xfs_attr_leaf.c
> +++ b/fs/xfs/libxfs/xfs_attr_leaf.c
> @@ -464,7 +464,7 @@ xfs_attr_copy_value(
>  	/*
>  	 * No copy if all we have to do is get the length
>  	 */
> -	if (args->flags & ATTR_KERNOVAL) {
> +	if (args->name.type & ATTR_KERNOVAL) {
>  		args->valuelen = valuelen;
>  		return 0;
>  	}
> @@ -679,27 +679,27 @@ xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff)
>  	sfe = &sf->list[0];
>  	for (i = 0; i < sf->hdr.count; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
>  #ifdef DEBUG
> -		if (sfe->namelen != args->namelen)
> +		if (sfe->namelen != args->name.len)
>  			continue;
> -		if (memcmp(args->name, sfe->nameval, args->namelen) != 0)
> +		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
>  			continue;
> -		if (!xfs_attr_namesp_match(args->flags, sfe->flags))
> +		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
>  			continue;
>  		ASSERT(0);
>  #endif
>  	}
>  
>  	offset = (char *)sfe - (char *)sf;
> -	size = XFS_ATTR_SF_ENTSIZE_BYNAME(args->namelen, args->valuelen);
> +	size = XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
>  	xfs_idata_realloc(dp, size, XFS_ATTR_FORK);
>  	sf = (xfs_attr_shortform_t *)ifp->if_u1.if_data;
>  	sfe = (xfs_attr_sf_entry_t *)((char *)sf + offset);
>  
> -	sfe->namelen = args->namelen;
> +	sfe->namelen = args->name.len;
>  	sfe->valuelen = args->valuelen;
> -	sfe->flags = XFS_ATTR_NSP_ARGS_TO_ONDISK(args->flags);
> -	memcpy(sfe->nameval, args->name, args->namelen);
> -	memcpy(&sfe->nameval[args->namelen], args->value, args->valuelen);
> +	sfe->flags = XFS_ATTR_NSP_ARGS_TO_ONDISK(args->name.type);
> +	memcpy(sfe->nameval, args->name.name, args->name.len);
> +	memcpy(&sfe->nameval[args->name.len], args->value, args->valuelen);
>  	sf->hdr.count++;
>  	be16_add_cpu(&sf->hdr.totsize, size);
>  	xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE | XFS_ILOG_ADATA);
> @@ -749,11 +749,11 @@ xfs_attr_shortform_remove(xfs_da_args_t *args)
>  	for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe),
>  					base += size, i++) {
>  		size = XFS_ATTR_SF_ENTSIZE(sfe);
> -		if (sfe->namelen != args->namelen)
> +		if (sfe->namelen != args->name.len)
>  			continue;
> -		if (memcmp(sfe->nameval, args->name, args->namelen) != 0)
> +		if (memcmp(sfe->nameval, args->name.name, args->name.len) != 0)
>  			continue;
> -		if (!xfs_attr_namesp_match(args->flags, sfe->flags))
> +		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
>  			continue;
>  		break;
>  	}
> @@ -816,11 +816,11 @@ xfs_attr_shortform_lookup(xfs_da_args_t *args)
>  	sfe = &sf->list[0];
>  	for (i = 0; i < sf->hdr.count;
>  				sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
> -		if (sfe->namelen != args->namelen)
> +		if (sfe->namelen != args->name.len)
>  			continue;
> -		if (memcmp(args->name, sfe->nameval, args->namelen) != 0)
> +		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
>  			continue;
> -		if (!xfs_attr_namesp_match(args->flags, sfe->flags))
> +		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
>  			continue;
>  		return -EEXIST;
>  	}
> @@ -847,13 +847,13 @@ xfs_attr_shortform_getvalue(
>  	sfe = &sf->list[0];
>  	for (i = 0; i < sf->hdr.count;
>  				sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
> -		if (sfe->namelen != args->namelen)
> +		if (sfe->namelen != args->name.len)
>  			continue;
> -		if (memcmp(args->name, sfe->nameval, args->namelen) != 0)
> +		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
>  			continue;
> -		if (!xfs_attr_namesp_match(args->flags, sfe->flags))
> +		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
>  			continue;
> -		return xfs_attr_copy_value(args, &sfe->nameval[args->namelen],
> +		return xfs_attr_copy_value(args, &sfe->nameval[args->name.len],
>  						sfe->valuelen);
>  	}
>  	return -ENOATTR;
> @@ -912,13 +912,13 @@ xfs_attr_shortform_to_leaf(
>  
>  	sfe = &sf->list[0];
>  	for (i = 0; i < sf->hdr.count; i++) {
> -		nargs.name = sfe->nameval;
> -		nargs.namelen = sfe->namelen;
> -		nargs.value = &sfe->nameval[nargs.namelen];
> +		nargs.name.name = sfe->nameval;
> +		nargs.name.len = sfe->namelen;
> +		nargs.value = &sfe->nameval[nargs.name.len];
>  		nargs.valuelen = sfe->valuelen;
>  		nargs.hashval = xfs_da_hashname(sfe->nameval,
>  						sfe->namelen);
> -		nargs.flags = XFS_ATTR_NSP_ONDISK_TO_ARGS(sfe->flags);
> +		nargs.name.type = XFS_ATTR_NSP_ONDISK_TO_ARGS(sfe->flags);
>  		error = xfs_attr3_leaf_lookup_int(bp, &nargs); /* set a->index */
>  		ASSERT(error == -ENOATTR);
>  		error = xfs_attr3_leaf_add(bp, &nargs);
> @@ -1119,12 +1119,12 @@ xfs_attr3_leaf_to_shortform(
>  			continue;
>  		ASSERT(entry->flags & XFS_ATTR_LOCAL);
>  		name_loc = xfs_attr3_leaf_name_local(leaf, i);
> -		nargs.name = name_loc->nameval;
> -		nargs.namelen = name_loc->namelen;
> -		nargs.value = &name_loc->nameval[nargs.namelen];
> +		nargs.name.name = name_loc->nameval;
> +		nargs.name.len = name_loc->namelen;
> +		nargs.value = &name_loc->nameval[nargs.name.len];
>  		nargs.valuelen = be16_to_cpu(name_loc->valuelen);
>  		nargs.hashval = be32_to_cpu(entry->hashval);
> -		nargs.flags = XFS_ATTR_NSP_ONDISK_TO_ARGS(entry->flags);
> +		nargs.name.type = XFS_ATTR_NSP_ONDISK_TO_ARGS(entry->flags);
>  		xfs_attr_shortform_add(&nargs, forkoff);
>  	}
>  	error = 0;
> @@ -1450,7 +1450,7 @@ xfs_attr3_leaf_add_work(
>  				     ichdr->freemap[mapindex].size);
>  	entry->hashval = cpu_to_be32(args->hashval);
>  	entry->flags = tmp ? XFS_ATTR_LOCAL : 0;
> -	entry->flags |= XFS_ATTR_NSP_ARGS_TO_ONDISK(args->flags);
> +	entry->flags |= XFS_ATTR_NSP_ARGS_TO_ONDISK(args->name.type);
>  	if (args->op_flags & XFS_DA_OP_RENAME) {
>  		entry->flags |= XFS_ATTR_INCOMPLETE;
>  		if ((args->blkno2 == args->blkno) &&
> @@ -1474,15 +1474,16 @@ xfs_attr3_leaf_add_work(
>  	 */
>  	if (entry->flags & XFS_ATTR_LOCAL) {
>  		name_loc = xfs_attr3_leaf_name_local(leaf, args->index);
> -		name_loc->namelen = args->namelen;
> +		name_loc->namelen = args->name.len;
>  		name_loc->valuelen = cpu_to_be16(args->valuelen);
> -		memcpy((char *)name_loc->nameval, args->name, args->namelen);
> -		memcpy((char *)&name_loc->nameval[args->namelen], args->value,
> +		memcpy((char *)name_loc->nameval, args->name.name,
> +		       args->name.len);
> +		memcpy((char *)&name_loc->nameval[args->name.len], args->value,
>  				   be16_to_cpu(name_loc->valuelen));
>  	} else {
>  		name_rmt = xfs_attr3_leaf_name_remote(leaf, args->index);
> -		name_rmt->namelen = args->namelen;
> -		memcpy((char *)name_rmt->name, args->name, args->namelen);
> +		name_rmt->namelen = args->name.len;
> +		memcpy((char *)name_rmt->name, args->name.name, args->name.len);
>  		entry->flags |= XFS_ATTR_INCOMPLETE;
>  		/* just in case */
>  		name_rmt->valuelen = 0;
> @@ -2409,23 +2410,25 @@ xfs_attr3_leaf_lookup_int(
>  		}
>  		if (entry->flags & XFS_ATTR_LOCAL) {
>  			name_loc = xfs_attr3_leaf_name_local(leaf, probe);
> -			if (name_loc->namelen != args->namelen)
> +			if (name_loc->namelen != args->name.len)
>  				continue;
> -			if (memcmp(args->name, name_loc->nameval,
> -							args->namelen) != 0)
> +			if (memcmp(args->name.name, name_loc->nameval,
> +							args->name.len) != 0)
>  				continue;
> -			if (!xfs_attr_namesp_match(args->flags, entry->flags))
> +			if (!xfs_attr_namesp_match(args->name.type,
> +						   entry->flags))
>  				continue;
>  			args->index = probe;
>  			return -EEXIST;
>  		} else {
>  			name_rmt = xfs_attr3_leaf_name_remote(leaf, probe);
> -			if (name_rmt->namelen != args->namelen)
> +			if (name_rmt->namelen != args->name.len)
>  				continue;
> -			if (memcmp(args->name, name_rmt->name,
> -							args->namelen) != 0)
> +			if (memcmp(args->name.name, name_rmt->name,
> +							args->name.len) != 0)
>  				continue;
> -			if (!xfs_attr_namesp_match(args->flags, entry->flags))
> +			if (!xfs_attr_namesp_match(args->name.type,
> +						   entry->flags))
>  				continue;
>  			args->index = probe;
>  			args->rmtvaluelen = be32_to_cpu(name_rmt->valuelen);
> @@ -2467,16 +2470,17 @@ xfs_attr3_leaf_getvalue(
>  	entry = &xfs_attr3_leaf_entryp(leaf)[args->index];
>  	if (entry->flags & XFS_ATTR_LOCAL) {
>  		name_loc = xfs_attr3_leaf_name_local(leaf, args->index);
> -		ASSERT(name_loc->namelen == args->namelen);
> -		ASSERT(memcmp(args->name, name_loc->nameval, args->namelen) == 0);
> +		ASSERT(name_loc->namelen == args->name.len);
> +		ASSERT(memcmp(args->name.name, name_loc->nameval,
> +			      args->name.len) == 0);
>  		return xfs_attr_copy_value(args,
> -					&name_loc->nameval[args->namelen],
> +					&name_loc->nameval[args->name.len],
>  					be16_to_cpu(name_loc->valuelen));
>  	}
>  
>  	name_rmt = xfs_attr3_leaf_name_remote(leaf, args->index);
> -	ASSERT(name_rmt->namelen == args->namelen);
> -	ASSERT(memcmp(args->name, name_rmt->name, args->namelen) == 0);
> +	ASSERT(name_rmt->namelen == args->name.len);
> +	ASSERT(memcmp(args->name.name, name_rmt->name, args->name.len) == 0);
>  	args->rmtvaluelen = be32_to_cpu(name_rmt->valuelen);
>  	args->rmtblkno = be32_to_cpu(name_rmt->valueblk);
>  	args->rmtblkcnt = xfs_attr3_rmt_blocks(args->dp->i_mount,
> @@ -2692,7 +2696,7 @@ xfs_attr_leaf_newentsize(
>  {
>  	int			size;
>  
> -	size = xfs_attr_leaf_entsize_local(args->namelen, args->valuelen);
> +	size = xfs_attr_leaf_entsize_local(args->name.len, args->valuelen);
>  	if (size < xfs_attr_leaf_entsize_local_max(args->geo->blksize)) {
>  		if (local)
>  			*local = 1;
> @@ -2700,7 +2704,7 @@ xfs_attr_leaf_newentsize(
>  	}
>  	if (local)
>  		*local = 0;
> -	return xfs_attr_leaf_entsize_remote(args->namelen);
> +	return xfs_attr_leaf_entsize_remote(args->name.len);
>  }
>  
>  
> @@ -2754,8 +2758,8 @@ xfs_attr3_leaf_clearflag(
>  		name = (char *)name_rmt->name;
>  	}
>  	ASSERT(be32_to_cpu(entry->hashval) == args->hashval);
> -	ASSERT(namelen == args->namelen);
> -	ASSERT(memcmp(name, args->name, namelen) == 0);
> +	ASSERT(namelen == args->name.len);
> +	ASSERT(memcmp(name, args->name.name, namelen) == 0);
>  #endif /* DEBUG */
>  
>  	entry->flags &= ~XFS_ATTR_INCOMPLETE;
> diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
> index a6ef5df..3e725c3 100644
> --- a/fs/xfs/libxfs/xfs_attr_remote.c
> +++ b/fs/xfs/libxfs/xfs_attr_remote.c
> @@ -380,7 +380,7 @@ xfs_attr_rmtval_get(
>  
>  	trace_xfs_attr_rmtval_get(args);
>  
> -	ASSERT(!(args->flags & ATTR_KERNOVAL));
> +	ASSERT(!(args->name.type & ATTR_KERNOVAL));
>  	ASSERT(args->rmtvaluelen == args->valuelen);
>  
>  	valuelen = args->rmtvaluelen;
> diff --git a/fs/xfs/libxfs/xfs_da_btree.c b/fs/xfs/libxfs/xfs_da_btree.c
> index 8c3eafe..da7e1af 100644
> --- a/fs/xfs/libxfs/xfs_da_btree.c
> +++ b/fs/xfs/libxfs/xfs_da_btree.c
> @@ -2125,8 +2125,10 @@ xfs_da_compname(
>  	const unsigned char *name,
>  	int		len)
>  {
> -	return (args->namelen == len && memcmp(args->name, name, len) == 0) ?
> -					XFS_CMP_EXACT : XFS_CMP_DIFFERENT;
> +	if (args->name.len == len && !memcmp(args->name.name, name, len))
> +		return XFS_CMP_EXACT;
> +
> +	return XFS_CMP_DIFFERENT;
>  }
>  
>  int
> diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
> index 0f4fbb0..14f1be3 100644
> --- a/fs/xfs/libxfs/xfs_da_btree.h
> +++ b/fs/xfs/libxfs/xfs_da_btree.h
> @@ -54,12 +54,10 @@ enum xfs_dacmp {
>   */
>  typedef struct xfs_da_args {
>  	struct xfs_da_geometry *geo;	/* da block geometry */
> -	const uint8_t		*name;		/* string (maybe not NULL terminated) */
> -	int		namelen;	/* length of string (maybe no NULL) */
> +	struct xfs_name	name;		/* name, length and argument  flags*/
>  	uint8_t		filetype;	/* filetype of inode for directories */
>  	uint8_t		*value;		/* set of bytes (maybe contain NULLs) */
>  	int		valuelen;	/* length of value */
> -	int		flags;		/* argument flags (eg: ATTR_NOCREATE) */
>  	xfs_dahash_t	hashval;	/* hash value of name */
>  	xfs_ino_t	inumber;	/* input/output inode number */
>  	struct xfs_inode *dp;		/* directory inode to manipulate */
> diff --git a/fs/xfs/libxfs/xfs_dir2.c b/fs/xfs/libxfs/xfs_dir2.c
> index dd6fcaa..3aadddc 100644
> --- a/fs/xfs/libxfs/xfs_dir2.c
> +++ b/fs/xfs/libxfs/xfs_dir2.c
> @@ -74,14 +74,14 @@ xfs_ascii_ci_compname(
>  	enum xfs_dacmp		result;
>  	int			i;
>  
> -	if (args->namelen != len)
> +	if (args->name.len != len)
>  		return XFS_CMP_DIFFERENT;
>  
>  	result = XFS_CMP_EXACT;
>  	for (i = 0; i < len; i++) {
> -		if (args->name[i] == name[i])
> +		if (args->name.name[i] == name[i])
>  			continue;
> -		if (tolower(args->name[i]) != tolower(name[i]))
> +		if (tolower(args->name.name[i]) != tolower(name[i]))
>  			return XFS_CMP_DIFFERENT;
>  		result = XFS_CMP_CASE;
>  	}
> @@ -265,8 +265,7 @@ xfs_dir_createname(
>  		return -ENOMEM;
>  
>  	args->geo = dp->i_mount->m_dir_geo;
> -	args->name = name->name;
> -	args->namelen = name->len;
> +	memcpy(&args->name, name, sizeof(struct xfs_name));
>  	args->filetype = name->type;
>  	args->hashval = xfs_dir2_hashname(dp->i_mount, name);
>  	args->inumber = inum;
> @@ -361,8 +360,7 @@ xfs_dir_lookup(
>  	 */
>  	args = kmem_zalloc(sizeof(*args), KM_NOFS);
>  	args->geo = dp->i_mount->m_dir_geo;
> -	args->name = name->name;
> -	args->namelen = name->len;
> +	memcpy(&args->name, name, sizeof(struct xfs_name));
>  	args->filetype = name->type;
>  	args->hashval = xfs_dir2_hashname(dp->i_mount, name);
>  	args->dp = dp;
> @@ -433,8 +431,7 @@ xfs_dir_removename(
>  		return -ENOMEM;
>  
>  	args->geo = dp->i_mount->m_dir_geo;
> -	args->name = name->name;
> -	args->namelen = name->len;
> +	memcpy(&args->name, name, sizeof(struct xfs_name));
>  	args->filetype = name->type;
>  	args->hashval = xfs_dir2_hashname(dp->i_mount, name);
>  	args->inumber = ino;
> @@ -494,8 +491,7 @@ xfs_dir_replace(
>  		return -ENOMEM;
>  
>  	args->geo = dp->i_mount->m_dir_geo;
> -	args->name = name->name;
> -	args->namelen = name->len;
> +	memcpy(&args->name, name, sizeof(struct xfs_name));
>  	args->filetype = name->type;
>  	args->hashval = xfs_dir2_hashname(dp->i_mount, name);
>  	args->inumber = inum;
> diff --git a/fs/xfs/libxfs/xfs_dir2_block.c b/fs/xfs/libxfs/xfs_dir2_block.c
> index d6ced59..592e47c 100644
> --- a/fs/xfs/libxfs/xfs_dir2_block.c
> +++ b/fs/xfs/libxfs/xfs_dir2_block.c
> @@ -355,7 +355,7 @@ xfs_dir2_block_addname(
>  	if (error)
>  		return error;
>  
> -	len = xfs_dir2_data_entsize(dp->i_mount, args->namelen);
> +	len = xfs_dir2_data_entsize(dp->i_mount, args->name.len);
>  
>  	/*
>  	 * Set up pointers to parts of the block.
> @@ -539,8 +539,8 @@ xfs_dir2_block_addname(
>  	 * Create the new data entry.
>  	 */
>  	dep->inumber = cpu_to_be64(args->inumber);
> -	dep->namelen = args->namelen;
> -	memcpy(dep->name, args->name, args->namelen);
> +	dep->namelen = args->name.len;
> +	memcpy(dep->name, args->name.name, args->name.len);
>  	xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
>  	tagp = xfs_dir2_data_entry_tag_p(dp->i_mount, dep);
>  	*tagp = cpu_to_be16((char *)dep - (char *)hdr);
> diff --git a/fs/xfs/libxfs/xfs_dir2_leaf.c b/fs/xfs/libxfs/xfs_dir2_leaf.c
> index a131b52..24a7fda 100644
> --- a/fs/xfs/libxfs/xfs_dir2_leaf.c
> +++ b/fs/xfs/libxfs/xfs_dir2_leaf.c
> @@ -653,7 +653,7 @@ xfs_dir2_leaf_addname(
>  	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
>  	ents = leafhdr.ents;
>  	bestsp = xfs_dir2_leaf_bests_p(ltp);
> -	length = xfs_dir2_data_entsize(dp->i_mount, args->namelen);
> +	length = xfs_dir2_data_entsize(dp->i_mount, args->name.len);
>  
>  	/*
>  	 * See if there are any entries with the same hash value
> @@ -856,8 +856,8 @@ xfs_dir2_leaf_addname(
>  	 */
>  	dep = (xfs_dir2_data_entry_t *)dup;
>  	dep->inumber = cpu_to_be64(args->inumber);
> -	dep->namelen = args->namelen;
> -	memcpy(dep->name, args->name, dep->namelen);
> +	dep->namelen = args->name.len;
> +	memcpy(dep->name, args->name.name, dep->namelen);
>  	xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
>  	tagp = xfs_dir2_data_entry_tag_p(dp->i_mount, dep);
>  	*tagp = cpu_to_be16((char *)dep - (char *)hdr);
> diff --git a/fs/xfs/libxfs/xfs_dir2_node.c b/fs/xfs/libxfs/xfs_dir2_node.c
> index a0cc5e2..3c74efc 100644
> --- a/fs/xfs/libxfs/xfs_dir2_node.c
> +++ b/fs/xfs/libxfs/xfs_dir2_node.c
> @@ -666,7 +666,7 @@ xfs_dir2_leafn_lookup_for_addname(
>  		ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC) ||
>  		       free->hdr.magic == cpu_to_be32(XFS_DIR3_FREE_MAGIC));
>  	}
> -	length = xfs_dir2_data_entsize(mp, args->namelen);
> +	length = xfs_dir2_data_entsize(mp, args->name.len);
>  	/*
>  	 * Loop over leaf entries with the right hash value.
>  	 */
> @@ -1911,7 +1911,7 @@ xfs_dir2_node_addname_int(
>  	int			needscan = 0;	/* need to rescan data frees */
>  	__be16			*tagp;		/* data entry tag pointer */
>  
> -	length = xfs_dir2_data_entsize(dp->i_mount, args->namelen);
> +	length = xfs_dir2_data_entsize(dp->i_mount, args->name.len);
>  	error = xfs_dir2_node_find_freeblk(args, fblk, &dbno, &fbp, &freehdr,
>  					   &findex, length);
>  	if (error)
> @@ -1966,8 +1966,8 @@ xfs_dir2_node_addname_int(
>  	/* Fill in the new entry and log it. */
>  	dep = (xfs_dir2_data_entry_t *)dup;
>  	dep->inumber = cpu_to_be64(args->inumber);
> -	dep->namelen = args->namelen;
> -	memcpy(dep->name, args->name, dep->namelen);
> +	dep->namelen = args->name.len;
> +	memcpy(dep->name, args->name.name, dep->namelen);
>  	xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
>  	tagp = xfs_dir2_data_entry_tag_p(dp->i_mount, dep);
>  	*tagp = cpu_to_be16((char *)dep - (char *)hdr);
> diff --git a/fs/xfs/libxfs/xfs_dir2_sf.c b/fs/xfs/libxfs/xfs_dir2_sf.c
> index 7b7f6fb..058c526 100644
> --- a/fs/xfs/libxfs/xfs_dir2_sf.c
> +++ b/fs/xfs/libxfs/xfs_dir2_sf.c
> @@ -387,7 +387,7 @@ xfs_dir2_sf_addname(
>  	/*
>  	 * Compute entry (and change in) size.
>  	 */
> -	incr_isize = xfs_dir2_sf_entsize(dp->i_mount, sfp, args->namelen);
> +	incr_isize = xfs_dir2_sf_entsize(dp->i_mount, sfp, args->name.len);
>  	objchange = 0;
>  
>  	/*
> @@ -470,7 +470,7 @@ xfs_dir2_sf_addname_easy(
>  	/*
>  	 * Grow the in-inode space.
>  	 */
> -	xfs_idata_realloc(dp, xfs_dir2_sf_entsize(mp, sfp, args->namelen),
> +	xfs_idata_realloc(dp, xfs_dir2_sf_entsize(mp, sfp, args->name.len),
>  			  XFS_DATA_FORK);
>  	/*
>  	 * Need to set up again due to realloc of the inode data.
> @@ -480,9 +480,9 @@ xfs_dir2_sf_addname_easy(
>  	/*
>  	 * Fill in the new entry.
>  	 */
> -	sfep->namelen = args->namelen;
> +	sfep->namelen = args->name.len;
>  	xfs_dir2_sf_put_offset(sfep, offset);
> -	memcpy(sfep->name, args->name, sfep->namelen);
> +	memcpy(sfep->name, args->name.name, sfep->namelen);
>  	xfs_dir2_sf_put_ino(mp, sfp, sfep, args->inumber);
>  	xfs_dir2_sf_put_ftype(mp, sfep, args->filetype);
>  
> @@ -540,7 +540,7 @@ xfs_dir2_sf_addname_hard(
>  	 */
>  	for (offset = args->geo->data_first_offset,
>  	      oldsfep = xfs_dir2_sf_firstentry(oldsfp),
> -	      add_datasize = xfs_dir2_data_entsize(mp, args->namelen),
> +	      add_datasize = xfs_dir2_data_entsize(mp, args->name.len),
>  	      eof = (char *)oldsfep == &buf[old_isize];
>  	     !eof;
>  	     offset = new_offset + xfs_dir2_data_entsize(mp, oldsfep->namelen),
> @@ -570,9 +570,9 @@ xfs_dir2_sf_addname_hard(
>  	/*
>  	 * Fill in the new entry, and update the header counts.
>  	 */
> -	sfep->namelen = args->namelen;
> +	sfep->namelen = args->name.len;
>  	xfs_dir2_sf_put_offset(sfep, offset);
> -	memcpy(sfep->name, args->name, sfep->namelen);
> +	memcpy(sfep->name, args->name.name, sfep->namelen);
>  	xfs_dir2_sf_put_ino(mp, sfp, sfep, args->inumber);
>  	xfs_dir2_sf_put_ftype(mp, sfep, args->filetype);
>  	sfp->count++;
> @@ -615,7 +615,7 @@ xfs_dir2_sf_addname_pick(
>  	int			used;		/* data bytes used */
>  
>  	sfp = (xfs_dir2_sf_hdr_t *)dp->i_df.if_u1.if_data;
> -	size = xfs_dir2_data_entsize(mp, args->namelen);
> +	size = xfs_dir2_data_entsize(mp, args->name.len);
>  	offset = args->geo->data_first_offset;
>  	sfep = xfs_dir2_sf_firstentry(sfp);
>  	holefit = 0;
> @@ -887,7 +887,7 @@ xfs_dir2_sf_lookup(
>  	/*
>  	 * Special case for .
>  	 */
> -	if (args->namelen == 1 && args->name[0] == '.') {
> +	if (args->name.len == 1 && args->name.name[0] == '.') {
>  		args->inumber = dp->i_ino;
>  		args->cmpresult = XFS_CMP_EXACT;
>  		args->filetype = XFS_DIR3_FT_DIR;
> @@ -896,8 +896,8 @@ xfs_dir2_sf_lookup(
>  	/*
>  	 * Special case for ..
>  	 */
> -	if (args->namelen == 2 &&
> -	    args->name[0] == '.' && args->name[1] == '.') {
> +	if (args->name.len == 2 &&
> +	    args->name.name[0] == '.' && args->name.name[1] == '.') {
>  		args->inumber = xfs_dir2_sf_get_parent_ino(sfp);
>  		args->cmpresult = XFS_CMP_EXACT;
>  		args->filetype = XFS_DIR3_FT_DIR;
> @@ -984,7 +984,7 @@ xfs_dir2_sf_removename(
>  	 * Calculate sizes.
>  	 */
>  	byteoff = (int)((char *)sfep - (char *)sfp);
> -	entsize = xfs_dir2_sf_entsize(mp, sfp, args->namelen);
> +	entsize = xfs_dir2_sf_entsize(mp, sfp, args->name.len);
>  	newsize = oldsize - entsize;
>  	/*
>  	 * Copy the part if any after the removed entry, sliding it down.
> @@ -1085,12 +1085,12 @@ xfs_dir2_sf_replace(
>  	} else
>  		i8elevated = 0;
>  
> -	ASSERT(args->namelen != 1 || args->name[0] != '.');
> +	ASSERT(args->name.len != 1 || args->name.name[0] != '.');
>  	/*
>  	 * Replace ..'s entry.
>  	 */
> -	if (args->namelen == 2 &&
> -	    args->name[0] == '.' && args->name[1] == '.') {
> +	if (args->name.len == 2 &&
> +	    args->name.name[0] == '.' && args->name.name[1] == '.') {
>  		ino = xfs_dir2_sf_get_parent_ino(sfp);
>  		ASSERT(args->inumber != ino);
>  		xfs_dir2_sf_put_parent_ino(sfp, args->inumber);
> diff --git a/fs/xfs/scrub/attr.c b/fs/xfs/scrub/attr.c
> index d9f0dd4..d4a9fe4 100644
> --- a/fs/xfs/scrub/attr.c
> +++ b/fs/xfs/scrub/attr.c
> @@ -147,17 +147,17 @@ xchk_xattr_listent(
>  		return;
>  	}
>  
> -	args.flags = ATTR_KERNOTIME;
> +	args.name.type = ATTR_KERNOTIME;
>  	if (flags & XFS_ATTR_ROOT)
> -		args.flags |= ATTR_ROOT;
> +		args.name.type |= ATTR_ROOT;
>  	else if (flags & XFS_ATTR_SECURE)
> -		args.flags |= ATTR_SECURE;
> +		args.name.type |= ATTR_SECURE;
>  	args.geo = context->dp->i_mount->m_attr_geo;
>  	args.whichfork = XFS_ATTR_FORK;
>  	args.dp = context->dp;
> -	args.name = name;
> -	args.namelen = namelen;
> -	args.hashval = xfs_da_hashname(args.name, args.namelen);
> +	args.name.name = name;
> +	args.name.len = namelen;
> +	args.hashval = xfs_da_hashname(args.name.name, args.name.len);
>  	args.trans = context->tp;
>  	args.value = xchk_xattr_valuebuf(sx->sc);
>  	args.valuelen = valuelen;
> diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
> index a86be7f..159b8af 100644
> --- a/fs/xfs/xfs_trace.h
> +++ b/fs/xfs/xfs_trace.h
> @@ -1633,7 +1633,7 @@ DECLARE_EVENT_CLASS(xfs_da_class,
>  	TP_STRUCT__entry(
>  		__field(dev_t, dev)
>  		__field(xfs_ino_t, ino)
> -		__dynamic_array(char, name, args->namelen)
> +		__dynamic_array(char, name, args->name.len)
>  		__field(int, namelen)
>  		__field(xfs_dahash_t, hashval)
>  		__field(xfs_ino_t, inumber)
> @@ -1642,9 +1642,10 @@ DECLARE_EVENT_CLASS(xfs_da_class,
>  	TP_fast_assign(
>  		__entry->dev = VFS_I(args->dp)->i_sb->s_dev;
>  		__entry->ino = args->dp->i_ino;
> -		if (args->namelen)
> -			memcpy(__get_str(name), args->name, args->namelen);
> -		__entry->namelen = args->namelen;
> +		if (args->name.len)
> +			memcpy(__get_str(name), args->name.name,
> +			       args->name.len);
> +		__entry->namelen = args->name.len;
>  		__entry->hashval = args->hashval;
>  		__entry->inumber = args->inumber;
>  		__entry->op_flags = args->op_flags;
> @@ -1697,7 +1698,7 @@ DECLARE_EVENT_CLASS(xfs_attr_class,
>  	TP_STRUCT__entry(
>  		__field(dev_t, dev)
>  		__field(xfs_ino_t, ino)
> -		__dynamic_array(char, name, args->namelen)
> +		__dynamic_array(char, name, args->name.len)
>  		__field(int, namelen)
>  		__field(int, valuelen)
>  		__field(xfs_dahash_t, hashval)
> @@ -1707,12 +1708,13 @@ DECLARE_EVENT_CLASS(xfs_attr_class,
>  	TP_fast_assign(
>  		__entry->dev = VFS_I(args->dp)->i_sb->s_dev;
>  		__entry->ino = args->dp->i_ino;
> -		if (args->namelen)
> -			memcpy(__get_str(name), args->name, args->namelen);
> -		__entry->namelen = args->namelen;
> +		if (args->name.len)
> +			memcpy(__get_str(name), args->name.name,
> +			       args->name.len);
> +		__entry->namelen = args->name.len;
>  		__entry->valuelen = args->valuelen;
>  		__entry->hashval = args->hashval;
> -		__entry->flags = args->flags;
> +		__entry->flags = args->name.type;
>  		__entry->op_flags = args->op_flags;
>  	),
>  	TP_printk("dev %d:%d ino 0x%llx name %.*s namelen %d valuelen %d "
> -- 
> 2.7.4
> 

Re: [PATCH v6 03/16] xfs: Add xfs_has_attr and subroutines

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:22PM -0700, Allison Collins wrote:
> From: Allison Henderson <allison.henderson@oracle.com>
> 
> This patch adds a new functions to check for the existence of an
> attribute.  Subroutines are also added to handle the cases of leaf
> blocks, nodes or shortform.  Common code that appears in existing attr
> add and remove functions have been factored out to help reduce the
> appearance of duplicated code.  We will need these routines later for
> delayed attributes since delayed operations cannot return error codes.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c      | 171 ++++++++++++++++++++++++++++--------------
>  fs/xfs/libxfs/xfs_attr.h      |   1 +
>  fs/xfs/libxfs/xfs_attr_leaf.c | 111 +++++++++++++++++----------
>  fs/xfs/libxfs/xfs_attr_leaf.h |   3 +
>  4 files changed, 188 insertions(+), 98 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index 9acdb23..a2507b9 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -46,6 +46,7 @@ STATIC int xfs_attr_shortform_addname(xfs_da_args_t *args);
>  STATIC int xfs_attr_leaf_get(xfs_da_args_t *args);
>  STATIC int xfs_attr_leaf_addname(xfs_da_args_t *args);
>  STATIC int xfs_attr_leaf_removename(xfs_da_args_t *args);
> +STATIC int xfs_attr_leaf_hasname(struct xfs_da_args *args, struct xfs_buf **bp);
>  
>  /*
>   * Internal routines when attribute list is more than one block.
> @@ -53,6 +54,8 @@ STATIC int xfs_attr_leaf_removename(xfs_da_args_t *args);
>  STATIC int xfs_attr_node_get(xfs_da_args_t *args);
>  STATIC int xfs_attr_node_addname(xfs_da_args_t *args);
>  STATIC int xfs_attr_node_removename(xfs_da_args_t *args);
> +STATIC int xfs_attr_node_hasname(xfs_da_args_t *args,
> +				 struct xfs_da_state **state);
>  STATIC int xfs_attr_fillstate(xfs_da_state_t *state);
>  STATIC int xfs_attr_refillstate(xfs_da_state_t *state);
>  
> @@ -310,6 +313,37 @@ xfs_attr_set_args(
>  }
>  
>  /*
> + * Return EEXIST if attr is found, or ENOATTR if not
> + */
> +int
> +xfs_has_attr(
> +	struct xfs_da_args      *args)
> +{
> +	struct xfs_inode	*dp = args->dp;
> +	struct xfs_buf		*bp = NULL;
> +	int			error;
> +
> +	if (!xfs_inode_hasattr(dp))
> +		return -ENOATTR;
> +
> +	if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
> +		ASSERT(dp->i_afp->if_flags & XFS_IFINLINE);
> +		return xfs_attr_sf_findname(args, NULL, NULL);
> +	}
> +
> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
> +		error = xfs_attr_leaf_hasname(args, &bp);
> +
> +		if (bp)
> +			xfs_trans_brelse(args->trans, bp);
> +
> +		return error;
> +	}
> +
> +	return xfs_attr_node_hasname(args, NULL);
> +}
> +
> +/*
>   * Remove the attribute specified in @args.
>   */
>  int
> @@ -583,26 +617,20 @@ STATIC int
>  xfs_attr_leaf_addname(
>  	struct xfs_da_args	*args)
>  {
> -	struct xfs_inode	*dp;
>  	struct xfs_buf		*bp;
>  	int			retval, error, forkoff;
> +	struct xfs_inode	*dp = args->dp;
>  
>  	trace_xfs_attr_leaf_addname(args);
>  
>  	/*
> -	 * Read the (only) block in the attribute list in.
> -	 */
> -	dp = args->dp;
> -	args->blkno = 0;

Don't we need this initialization?

> -	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
> -	if (error)
> -		return error;
> -
> -	/*
>  	 * Look up the given attribute in the leaf block.  Figure out if
>  	 * the given flags produce an error or call for an atomic rename.
>  	 */
> -	retval = xfs_attr3_leaf_lookup_int(bp, args);
> +	retval = xfs_attr_leaf_hasname(args, &bp);
> +	if (retval != -ENOATTR && retval != -EEXIST)
> +		return retval;
> +
>  	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
>  		xfs_trans_brelse(args->trans, bp);
>  		return retval;
> @@ -754,6 +782,27 @@ xfs_attr_leaf_addname(
>  }
>  
>  /*
> + * Return EEXIST if attr is found, or ENOATTR if not
> + */
> +STATIC int
> +xfs_attr_leaf_hasname(
> +	struct xfs_da_args      *args,
> +	struct xfs_buf		**bp)
> +{
> +	int                     error = 0;
> +
> +	error = xfs_attr3_leaf_read(args->trans, args->dp, 0, bp);
> +	if (error)
> +		return error;
> +
> +	error = xfs_attr3_leaf_lookup_int(*bp, args);
> +	if (error != -ENOATTR && error != -EEXIST)
> +		xfs_trans_brelse(args->trans, *bp);
> +
> +	return error;
> +}
> +
> +/*
>   * Remove a name from the leaf attribute list structure
>   *
>   * This leaf block cannot have a "remote" value, we only call this routine
> @@ -773,12 +822,11 @@ xfs_attr_leaf_removename(
>  	 * Remove the attribute.
>  	 */
>  	dp = args->dp;
> -	args->blkno = 0;
> -	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
> -	if (error)
> +
> +	error = xfs_attr_leaf_hasname(args, &bp);
> +	if (error != -ENOATTR && error != -EEXIST)
>  		return error;
>  
> -	error = xfs_attr3_leaf_lookup_int(bp, args);
>  	if (error == -ENOATTR) {
>  		xfs_trans_brelse(args->trans, bp);
>  		return error;
> @@ -817,12 +865,10 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
>  
>  	trace_xfs_attr_leaf_get(args);
>  
> -	args->blkno = 0;
> -	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
> -	if (error)
> +	error = xfs_attr_leaf_hasname(args, &bp);
> +	if (error != -ENOATTR && error != -EEXIST)
>  		return error;
>  
> -	error = xfs_attr3_leaf_lookup_int(bp, args);
>  	if (error != -EEXIST)  {
>  		xfs_trans_brelse(args->trans, bp);
>  		return error;
> @@ -832,6 +878,41 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
>  	return error;
>  }
>  
> +/*
> + * Return EEXIST if attr is found, or ENOATTR if not
> + * statep: If not null is set to point at the found state.  Caller will
> + *         be responsible for freeing the state in this case.

Uh... a predicate returns an xfs_da_state object?

/me starts to wonder if these hasname functions ought to be some sort of
lookup function that returns the appropriate context for inserting a new
xattr entry if ENOATTR; a NULL context if EEXIST; or some other error if
something actually went wrong...?

Granted I have this funny feeling that most of the attr code works on
the principle that we're supposed to pass ENOATTR/EEXIST from lower
layers to higher layers, which makes me unsure if it makes more sense to
stop abusing those magic error codes (and therefore having to switch()
on every error return) or not.

I dunno, I also had questions that I attached to the previous version of
this patch, so... <shrug>

--D

> + */
> +STATIC int
> +xfs_attr_node_hasname(
> +	struct xfs_da_args	*args,
> +	struct xfs_da_state	**statep)
> +{
> +	struct xfs_da_state	*state;
> +	int			retval, error;
> +
> +	state = xfs_da_state_alloc();
> +	state->args = args;
> +	state->mp = args->dp->i_mount;
> +
> +	if (statep != NULL)
> +		*statep = NULL;
> +
> +	/*
> +	 * Search to see if name exists, and get back a pointer to it.
> +	 */
> +	error = xfs_da3_node_lookup_int(state, &retval);
> +	if (error == 0) {
> +		if (statep != NULL)
> +			*statep = state;
> +		return retval;
> +	}
> +
> +	xfs_da_state_free(state);
> +
> +	return error;
> +}
> +
>  /*========================================================================
>   * External routines when attribute list size > geo->blksize
>   *========================================================================*/
> @@ -864,20 +945,17 @@ xfs_attr_node_addname(
>  	dp = args->dp;
>  	mp = dp->i_mount;
>  restart:
> -	state = xfs_da_state_alloc();
> -	state->args = args;
> -	state->mp = mp;
> -
>  	/*
>  	 * Search to see if name already exists, and get back a pointer
>  	 * to where it should go.
>  	 */
> -	error = xfs_da3_node_lookup_int(state, &retval);
> -	if (error)
> +	retval = xfs_attr_node_hasname(args, &state);
> +	if (retval != -ENOATTR)
>  		goto out;
> +
>  	blk = &state->path.blk[ state->path.active-1 ];
>  	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
> -	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
> +	if (args->name.type & ATTR_REPLACE) {
>  		goto out;
>  	} else if (retval == -EEXIST) {
>  		if (args->name.type & ATTR_CREATE)
> @@ -1079,29 +1157,15 @@ xfs_attr_node_removename(
>  {
>  	struct xfs_da_state	*state;
>  	struct xfs_da_state_blk	*blk;
> -	struct xfs_inode	*dp;
>  	struct xfs_buf		*bp;
>  	int			retval, error, forkoff;
> +	struct xfs_inode	*dp = args->dp;
>  
>  	trace_xfs_attr_node_removename(args);
>  
> -	/*
> -	 * Tie a string around our finger to remind us where we are.
> -	 */
> -	dp = args->dp;
> -	state = xfs_da_state_alloc();
> -	state->args = args;
> -	state->mp = dp->i_mount;
> -
> -	/*
> -	 * Search to see if name exists, and get back a pointer to it.
> -	 */
> -	error = xfs_da3_node_lookup_int(state, &retval);
> -	if (error || (retval != -EEXIST)) {
> -		if (error == 0)
> -			error = retval;
> +	error = xfs_attr_node_hasname(args, &state);
> +	if (error != -EEXIST)
>  		goto out;
> -	}
>  
>  	/*
>  	 * If there is an out-of-line value, de-allocate the blocks.
> @@ -1196,7 +1260,8 @@ xfs_attr_node_removename(
>  	error = 0;
>  
>  out:
> -	xfs_da_state_free(state);
> +	if (state)
> +		xfs_da_state_free(state);
>  	return error;
>  }
>  
> @@ -1316,31 +1381,23 @@ xfs_attr_node_get(xfs_da_args_t *args)
>  {
>  	xfs_da_state_t *state;
>  	xfs_da_state_blk_t *blk;
> -	int error, retval;
> +	int error;
>  	int i;
>  
>  	trace_xfs_attr_node_get(args);
>  
> -	state = xfs_da_state_alloc();
> -	state->args = args;
> -	state->mp = args->dp->i_mount;
> -
>  	/*
>  	 * Search to see if name exists, and get back a pointer to it.
>  	 */
> -	error = xfs_da3_node_lookup_int(state, &retval);
> -	if (error) {
> -		retval = error;
> -		goto out_release;
> -	}
> -	if (retval != -EEXIST)
> +	error = xfs_attr_node_hasname(args, &state);
> +	if (error != -EEXIST)
>  		goto out_release;
>  
>  	/*
>  	 * Get the value, local or "remote"
>  	 */
>  	blk = &state->path.blk[state->path.active - 1];
> -	retval = xfs_attr3_leaf_getvalue(blk->bp, args);
> +	error = xfs_attr3_leaf_getvalue(blk->bp, args);
>  
>  	/*
>  	 * If not in a transaction, we have to release all the buffers.
> @@ -1352,7 +1409,7 @@ xfs_attr_node_get(xfs_da_args_t *args)
>  	}
>  
>  	xfs_da_state_free(state);
> -	return retval;
> +	return error;
>  }
>  
>  /* Returns true if the attribute entry name is valid. */
> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
> index 35043db..ce7b039 100644
> --- a/fs/xfs/libxfs/xfs_attr.h
> +++ b/fs/xfs/libxfs/xfs_attr.h
> @@ -153,6 +153,7 @@ int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
>  		 unsigned char *value, int valuelen, int flags);
>  int xfs_attr_set_args(struct xfs_da_args *args);
>  int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
> +int xfs_has_attr(struct xfs_da_args *args);
>  int xfs_attr_remove_args(struct xfs_da_args *args);
>  int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
>  		  int flags, struct attrlist_cursor_kern *cursor);
> diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
> index cb5ef66..9d6b68c 100644
> --- a/fs/xfs/libxfs/xfs_attr_leaf.c
> +++ b/fs/xfs/libxfs/xfs_attr_leaf.c
> @@ -654,18 +654,66 @@ xfs_attr_shortform_create(xfs_da_args_t *args)
>  }
>  
>  /*
> + * Return -EEXIST if attr is found, or -ENOATTR if not
> + * args:  args containing attribute name and namelen
> + * sfep:  If not null, pointer will be set to the last attr entry found on
> +	  -EEXIST.  On -ENOATTR pointer is left at the last entry in the list
> + * basep: If not null, pointer is set to the byte offset of the entry in the
> + *	  list on -EEXIST.  On -ENOATTR, pointer is left at the byte offset of
> + *	  the last entry in the list
> + */
> +int
> +xfs_attr_sf_findname(
> +	struct xfs_da_args	 *args,
> +	struct xfs_attr_sf_entry **sfep,
> +	unsigned int		 *basep)
> +{
> +	struct xfs_attr_shortform *sf;
> +	struct xfs_attr_sf_entry *sfe;
> +	unsigned int		base = sizeof(struct xfs_attr_sf_hdr);
> +	int			size = 0;
> +	int			end;
> +	int			i;
> +
> +	sf = (struct xfs_attr_shortform *)args->dp->i_afp->if_u1.if_data;
> +	sfe = &sf->list[0];
> +	end = sf->hdr.count;
> +	for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe),
> +			base += size, i++) {
> +		size = XFS_ATTR_SF_ENTSIZE(sfe);
> +		if (sfe->namelen != args->name.len)
> +			continue;
> +		if (memcmp(sfe->nameval, args->name.name, args->name.len) != 0)
> +			continue;
> +		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
> +			continue;
> +		break;
> +	}
> +
> +	if (sfep != NULL)
> +		*sfep = sfe;
> +
> +	if (basep != NULL)
> +		*basep = base;
> +
> +	if (i == end)
> +		return -ENOATTR;
> +	return -EEXIST;
> +}
> +
> +/*
>   * Add a name/value pair to the shortform attribute list.
>   * Overflow from the inode has already been checked for.
>   */
>  void
> -xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff)
> +xfs_attr_shortform_add(struct xfs_da_args *args, int forkoff)
>  {
> -	xfs_attr_shortform_t *sf;
> -	xfs_attr_sf_entry_t *sfe;
> -	int i, offset, size;
> -	xfs_mount_t *mp;
> -	xfs_inode_t *dp;
> -	struct xfs_ifork *ifp;
> +	struct xfs_attr_shortform	*sf;
> +	struct xfs_attr_sf_entry	*sfe;
> +	int				offset, size, error;
> +	struct xfs_mount		*mp;
> +	struct xfs_inode		*dp;
> +	struct xfs_ifork		*ifp;
>  
>  	trace_xfs_attr_sf_add(args);
>  
> @@ -676,18 +724,8 @@ xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff)
>  	ifp = dp->i_afp;
>  	ASSERT(ifp->if_flags & XFS_IFINLINE);
>  	sf = (xfs_attr_shortform_t *)ifp->if_u1.if_data;
> -	sfe = &sf->list[0];
> -	for (i = 0; i < sf->hdr.count; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
> -#ifdef DEBUG
> -		if (sfe->namelen != args->name.len)
> -			continue;
> -		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
> -			continue;
> -		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
> -			continue;
> -		ASSERT(0);
> -#endif
> -	}
> +	error = xfs_attr_sf_findname(args, &sfe, NULL);
> +	ASSERT(error != -EEXIST);
>  
>  	offset = (char *)sfe - (char *)sf;
>  	size = XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
> @@ -730,35 +768,26 @@ xfs_attr_fork_remove(
>   * Remove an attribute from the shortform attribute list structure.
>   */
>  int
> -xfs_attr_shortform_remove(xfs_da_args_t *args)
> +xfs_attr_shortform_remove(struct xfs_da_args *args)
>  {
> -	xfs_attr_shortform_t *sf;
> -	xfs_attr_sf_entry_t *sfe;
> -	int base, size=0, end, totsize, i;
> -	xfs_mount_t *mp;
> -	xfs_inode_t *dp;
> +	struct xfs_attr_shortform	*sf;
> +	struct xfs_attr_sf_entry	*sfe;
> +	int				size = 0, end, totsize;
> +	unsigned int			base;
> +	struct xfs_mount		*mp;
> +	struct xfs_inode		*dp;
> +	int				error;
>  
>  	trace_xfs_attr_sf_remove(args);
>  
>  	dp = args->dp;
>  	mp = dp->i_mount;
> -	base = sizeof(xfs_attr_sf_hdr_t);
>  	sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data;
> -	sfe = &sf->list[0];
> -	end = sf->hdr.count;
> -	for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe),
> -					base += size, i++) {
> -		size = XFS_ATTR_SF_ENTSIZE(sfe);
> -		if (sfe->namelen != args->name.len)
> -			continue;
> -		if (memcmp(sfe->nameval, args->name.name, args->name.len) != 0)
> -			continue;
> -		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
> -			continue;
> -		break;
> -	}
> -	if (i == end)
> -		return -ENOATTR;
> +
> +	error = xfs_attr_sf_findname(args, &sfe, &base);
> +	if (error != -EEXIST)
> +		return error;
> +	size = XFS_ATTR_SF_ENTSIZE(sfe);
>  
>  	/*
>  	 * Fix up the attribute fork data, covering the hole
> diff --git a/fs/xfs/libxfs/xfs_attr_leaf.h b/fs/xfs/libxfs/xfs_attr_leaf.h
> index f4a188e..f6165ff 100644
> --- a/fs/xfs/libxfs/xfs_attr_leaf.h
> +++ b/fs/xfs/libxfs/xfs_attr_leaf.h
> @@ -62,6 +62,9 @@ int	xfs_attr_shortform_getvalue(struct xfs_da_args *args);
>  int	xfs_attr_shortform_to_leaf(struct xfs_da_args *args,
>  			struct xfs_buf **leaf_bp);
>  int	xfs_attr_shortform_remove(struct xfs_da_args *args);
> +int	xfs_attr_sf_findname(struct xfs_da_args *args,
> +			     struct xfs_attr_sf_entry **sfep,
> +			     unsigned int *basep);
>  int	xfs_attr_shortform_allfit(struct xfs_buf *bp, struct xfs_inode *dp);
>  int	xfs_attr_shortform_bytesfit(struct xfs_inode *dp, int bytes);
>  xfs_failaddr_t xfs_attr_shortform_verify(struct xfs_inode *ip);
> -- 
> 2.7.4
> 

Re: [PATCH v6 06/16] xfs: Factor out xfs_attr_leaf_addname helper

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:25PM -0700, Allison Collins wrote:
> Factor out new helper function xfs_attr_leaf_try_add. Because new
> delayed attribute routines cannot roll transactions, we carve off the
> parts of xfs_attr_leaf_addname that we can use, and move the commit into
> the calling function.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> Reviewed-by: Brian Foster <bfoster@redhat.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c | 83 ++++++++++++++++++++++++++++--------------------
>  1 file changed, 49 insertions(+), 34 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index b0ec25b..9ed7e94 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -305,10 +305,30 @@ xfs_attr_set_args(
>  		}
>  	}
>  
> -	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK))
> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>  		error = xfs_attr_leaf_addname(args);
> -	else
> -		error = xfs_attr_node_addname(args);
> +		if (error != -ENOSPC)
> +			return error;
> +
> +		/*
> +		 * Commit that transaction so that the node_addname()
> +		 * call can manage its own transactions.
> +		 */
> +		error = xfs_defer_finish(&args->trans);
> +		if (error)
> +			return error;
> +
> +		/*
> +		 * Commit the current trans (including the inode) and
> +		 * start a new one.
> +		 */
> +		error = xfs_trans_roll_inode(&args->trans, dp);
> +		if (error)
> +			return error;
> +
> +	}
> +
> +	error = xfs_attr_node_addname(args);
>  	return error;
>  }
>  
> @@ -607,21 +627,12 @@ xfs_attr_shortform_addname(xfs_da_args_t *args)
>   * External routines when attribute list is one block
>   *========================================================================*/
>  
> -/*
> - * Add a name to the leaf attribute list structure
> - *
> - * This leaf block cannot have a "remote" value, we only call this routine
> - * if bmap_one_block() says there is only one block (ie: no remote blks).
> - */

These functions are complicated enough as they are now, please make sure
they all have comments laying out the expected input metadata states and
what output states result from them.

I /think/ this function takes an inode whose attr data is in leaf
format, and tries to add a new entry.  If that succeeds then we exit
with dirty inode and transaction.  If ENOSPC then we convert the attr
data to node format and exit w/ dirty inode + transaction, presuming
that the caller will try again with xfs_attr_node_addname?

>  STATIC int
> -xfs_attr_leaf_addname(
> -	struct xfs_da_args	*args)
> +xfs_attr_leaf_try_add(
> +	struct xfs_da_args	*args,
> +	struct xfs_buf		*bp)
>  {
> -	struct xfs_buf		*bp;
> -	int			retval, error, forkoff;
> -	struct xfs_inode	*dp = args->dp;
> -
> -	trace_xfs_attr_leaf_addname(args);
> +	int			retval, error;
>  
>  	/*
>  	 * Look up the given attribute in the leaf block.  Figure out if
> @@ -667,31 +678,35 @@ xfs_attr_leaf_addname(
>  	retval = xfs_attr3_leaf_add(bp, args);
>  	if (retval == -ENOSPC) {
>  		/*
> -		 * Promote the attribute list to the Btree format, then
> -		 * Commit that transaction so that the node_addname() call
> -		 * can manage its own transactions.
> +		 * Promote the attribute list to the Btree format.
> +		 * Unless an error occurs, retain the -ENOSPC retval
>  		 */
>  		error = xfs_attr3_leaf_to_node(args);
>  		if (error)
>  			return error;
> -		error = xfs_defer_finish(&args->trans);
> -		if (error)
> -			return error;
> +	}
> +	return retval;
> +}
>  
> -		/*
> -		 * Commit the current trans (including the inode) and start
> -		 * a new one.
> -		 */
> -		error = xfs_trans_roll_inode(&args->trans, dp);
> -		if (error)
> -			return error;
>  
> -		/*
> -		 * Fob the whole rest of the problem off on the Btree code.
> -		 */
> -		error = xfs_attr_node_addname(args);
> +/*
> + * Add a name to the leaf attribute list structure
> + *
> + * This leaf block cannot have a "remote" value, we only call this routine
> + * if bmap_one_block() says there is only one block (ie: no remote blks).
> + */
> +STATIC int
> +xfs_attr_leaf_addname(struct xfs_da_args	*args)

STATIC int
xfs_attr_leaf_addname(
	struct xfs_da_args	*args)
{

Please fix the inconsistent style things as you touch them.

--D

> +{
> +	int			error, forkoff;
> +	struct xfs_buf		*bp = NULL;
> +	struct xfs_inode	*dp = args->dp;
> +
> +	trace_xfs_attr_leaf_addname(args);
> +
> +	error = xfs_attr_leaf_try_add(args, bp);
> +	if (error)
>  		return error;
> -	}
>  
>  	/*
>  	 * Commit the transaction that added the attr name so that
> -- 
> 2.7.4
> 

Re: [PATCH v6 07/16] xfs: Refactor xfs_attr_try_sf_addname

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:26PM -0700, Allison Collins wrote:
> To help pre-simplify xfs_attr_set_args, we need to hoist transacation
> handling up, while modularizing the adjacent code down into helpers. In
> this patch, hoist the commit in xfs_attr_try_sf_addname up into the
> calling function, and also pull the attr list creation down.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c | 30 +++++++++++++++---------------
>  1 file changed, 15 insertions(+), 15 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index 9ed7e94..c15361a 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -227,8 +227,13 @@ xfs_attr_try_sf_addname(
>  	struct xfs_da_args	*args)
>  {
>  
> -	struct xfs_mount	*mp = dp->i_mount;
> -	int			error, error2;
> +	int			error;
> +
> +	/*
> +	 * Build initial attribute list (if required).
> +	 */
> +	if (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS)
> +		xfs_attr_shortform_create(args);
>  
>  	error = xfs_attr_shortform_addname(args);
>  	if (error == -ENOSPC)
> @@ -241,12 +246,10 @@ xfs_attr_try_sf_addname(
>  	if (!error && (args->name.type & ATTR_KERNOTIME) == 0)
>  		xfs_trans_ichgtime(args->trans, dp, XFS_ICHGTIME_CHG);
>  
> -	if (mp->m_flags & XFS_MOUNT_WSYNC)
> +	if (dp->i_mount->m_flags & XFS_MOUNT_WSYNC)
>  		xfs_trans_set_sync(args->trans);
>  
> -	error2 = xfs_trans_commit(args->trans);
> -	args->trans = NULL;
> -	return error ? error : error2;
> +	return error;
>  }
>  
>  /*
> @@ -258,7 +261,7 @@ xfs_attr_set_args(
>  {
>  	struct xfs_inode	*dp = args->dp;
>  	struct xfs_buf          *leaf_bp = NULL;
> -	int			error;
> +	int			error, error2 = 0;
>  
>  	/*
>  	 * If the attribute list is non-existent or a shortform list,
> @@ -269,17 +272,14 @@ xfs_attr_set_args(
>  	     dp->i_d.di_anextents == 0)) {
>  
>  		/*
> -		 * Build initial attribute list (if required).
> -		 */
> -		if (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS)
> -			xfs_attr_shortform_create(args);
> -
> -		/*
>  		 * Try to add the attr to the attribute list in the inode.
>  		 */
>  		error = xfs_attr_try_sf_addname(dp, args);
> -		if (error != -ENOSPC)
> -			return error;
> +		if (error != -ENOSPC) {
> +			error2 = xfs_trans_commit(args->trans);

/me wonders if there's really a point in committing the transaction for
things like EEXIST and ENOMEM, but I guess this is a straight
conversion and the current code really does this...

Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>

--D

> +			args->trans = NULL;
> +			return error ? error : error2;
> +		}
>  
>  		/*
>  		 * It won't fit in the shortform, transform to a leaf block.
> -- 
> 2.7.4
> 

Re: [PATCH v6 11/16] xfs: Check for -ENOATTR or -EEXIST

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:30PM -0700, Allison Collins wrote:
> Delayed operations cannot return error codes.  So we must check for
> these conditions first before starting set or remove operations

Answering my own question from earlier -- I see here you actually /are/
checking the attr existence w.r.t. ATTR_{CREATE,REPLACE} right after we
allocate a transaction and ILOCK the inode, so

Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>

Though I am wondering if you could discard the predicates from the
second patch in favor of doing a normal lookup of the attr with a zero
valuelen to determine if there's already an attribute?

--D

> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index a2673fe..e9d22c1 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -457,6 +457,14 @@ xfs_attr_set(
>  		goto out_trans_cancel;
>  
>  	xfs_trans_ijoin(args.trans, dp, 0);
> +
> +	error = xfs_has_attr(&args);
> +	if (error == -EEXIST && (name->type & ATTR_CREATE))
> +		goto out_trans_cancel;
> +
> +	if (error == -ENOATTR && (name->type & ATTR_REPLACE))
> +		goto out_trans_cancel;
> +
>  	error = xfs_attr_set_args(&args);
>  	if (error)
>  		goto out_trans_cancel;
> @@ -545,6 +553,10 @@ xfs_attr_remove(
>  	 */
>  	xfs_trans_ijoin(args.trans, dp, 0);
>  
> +	error = xfs_has_attr(&args);
> +	if (error != -EEXIST)
> +		goto out;
> +
>  	error = xfs_attr_remove_args(&args);
>  	if (error)
>  		goto out;
> -- 
> 2.7.4
> 

Re: [PATCH v6 12/16] xfs: Add helper function xfs_attr_init_unmapstate

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:31PM -0700, Allison Collins wrote:
> This patch helps to pre-simplify xfs_attr_node_removename by modularizing
> the code around the transactions into helper functions.  This will make
> the function easier to follow when we introduce delayed attributes.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c | 45 +++++++++++++++++++++++++++++++--------------
>  1 file changed, 31 insertions(+), 14 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index e9d22c1..453ea59 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -1202,6 +1202,36 @@ xfs_attr_node_addname(
>  }
>  
>  /*
> + * Set up the state for unmap and set the incomplete flag

"Mark an attribute entry INCOMPLETE and save pointers to the relevant
buffers for later deletion of the entry." ?

> + */
> +STATIC int
> +xfs_attr_init_unmapstate(

I'm hung up on this name -- it marks the entry incomplete and saves some
breadcrumbs for erasing data later, which I guess is "unmap state"?
What do you think of:

xfs_attr_leaf_mark_incomplete()

> +	struct xfs_da_args	*args,
> +	struct xfs_da_state	*state)
> +{
> +	int error;
> +
> +	/*
> +	 * Fill in disk block numbers in the state structure
> +	 * so that we can get the buffers back after we commit
> +	 * several transactions in the following calls.
> +	 */
> +	error = xfs_attr_fillstate(state);
> +	if (error)
> +		return error;
> +
> +	/*
> +	 * Mark the attribute as INCOMPLETE

Has this stopped doing the bunmapi call, or was the comment inaccurate?
I'm guessing the second if it's necessary to save state to recover
buffer pointers later...?

--D

> +	 */
> +	error = xfs_attr3_leaf_setflag(args);
> +	if (error)
> +		return error;
> +
> +	return 0;
> +}
> +
> +
> +/*
>   * Remove a name from a B-tree attribute list.
>   *
>   * This will involve walking down the Btree, and may involve joining
> @@ -1233,20 +1263,7 @@ xfs_attr_node_removename(
>  	ASSERT(blk->bp != NULL);
>  	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
>  	if (args->rmtblkno > 0) {
> -		/*
> -		 * Fill in disk block numbers in the state structure
> -		 * so that we can get the buffers back after we commit
> -		 * several transactions in the following calls.
> -		 */
> -		error = xfs_attr_fillstate(state);
> -		if (error)
> -			goto out;
> -
> -		/*
> -		 * Mark the attribute as INCOMPLETE, then bunmapi() the
> -		 * remote value.
> -		 */
> -		error = xfs_attr3_leaf_setflag(args);
> +		error = xfs_attr_init_unmapstate(args, state);
>  		if (error)
>  			goto out;
>  
> -- 
> 2.7.4
> 

Re: [PATCH v6 13/16] xfs: Add helper function xfs_attr_rmtval_unmap

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:32PM -0700, Allison Collins wrote:
> This function is similar to xfs_attr_rmtval_remove, but adapted to
> return. EAGAIN for new transactions. We will use this later when we

"..to return -EAGAIN"?

> introduce delayed attributes
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr_remote.c | 27 +++++++++++++++++++++++++++
>  fs/xfs/libxfs/xfs_attr_remote.h |  1 +
>  2 files changed, 28 insertions(+)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
> index 9b4fa2a..f2ee0b8 100644
> --- a/fs/xfs/libxfs/xfs_attr_remote.c
> +++ b/fs/xfs/libxfs/xfs_attr_remote.c
> @@ -684,3 +684,30 @@ xfs_attr_rmtval_remove(
>  	}
>  	return 0;
>  }
> +
> +/*
> + * Unmap value blocks for this attr.  This is similar to
> + * xfs_attr_rmtval_remove, but adapted to to return EAGAIN for new transactions

Do you have to scan for and invalidate any incore buffers for the remote
value?  Or will that be performed by another step?  Hard to tell because
this function doesn't have any callers.

> + */
> +int
> +xfs_attr_rmtval_unmap(
> +	struct xfs_da_args	*args)
> +{
> +	int	error, done;
> +
> +	/*
> +	 * Unmap value blocks for this attr.  This is similar to
> +	 * xfs_attr_rmtval_remove, but open coded here to return EAGAIN
> +	 * for new transactions
> +	 */
> +	error = xfs_bunmapi(args->trans, args->dp,
> +		    args->rmtblkno, args->rmtblkcnt,
> +		    XFS_BMAPI_ATTRFORK, 1, &done);
> +	if (error)
> +		return error;
> +
> +	if (!done)
> +		return -EAGAIN;

return done ? 0 : -EAGAIN; ?

--D

> +
> +	return 0;
> +}
> diff --git a/fs/xfs/libxfs/xfs_attr_remote.h b/fs/xfs/libxfs/xfs_attr_remote.h
> index 85f2593..7ab3770 100644
> --- a/fs/xfs/libxfs/xfs_attr_remote.h
> +++ b/fs/xfs/libxfs/xfs_attr_remote.h
> @@ -12,4 +12,5 @@ int xfs_attr_rmtval_get(struct xfs_da_args *args);
>  int xfs_attr_rmtval_set(struct xfs_da_args *args);
>  int xfs_attr_rmtval_remove(struct xfs_da_args *args);
>  int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
> +int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
>  #endif /* __XFS_ATTR_REMOTE_H__ */
> -- 
> 2.7.4
> 

Re: [PATCH v6 14/16] xfs: Simplify xfs_attr_set_args

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:33PM -0700, Allison Collins wrote:
> Delayed attribute mechanics make frequent use of goto statements.  We
> can use this to further simplify xfs_attr_set_args.  In this patch we
> introduce one of the gotos now to help pre-simplify the routine.  This
> will help make proceeding patches in this area easier to read.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c | 71 +++++++++++++++++++++++++-----------------------
>  1 file changed, 37 insertions(+), 34 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index 453ea59..90e0b2d 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -264,47 +264,50 @@ xfs_attr_set_args(
>  	int			error, error2 = 0;
>  
>  	/*
> -	 * If the attribute list is non-existent or a shortform list,
> -	 * upgrade it to a single-leaf-block attribute list.
> +	 * If the attribute list is non-existent or a shortform list, proceed to
> +	 * upgrade it to a single-leaf-block attribute list.  Otherwise jump
> +	 * straight to leaf handling

We don't jump to "upgrade it to a single leaf block attr list" quite
that soon.

"If the attribute list is already in leaf format, jump straight to leaf
handling.  Otherwise, try to add the attribute to the shortform list; if
there's no room then convert the list to leaf format and try again." ?

>  	 */
> -	if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL ||
> +	if (!(dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL ||
>  	    (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
> -	     dp->i_d.di_anextents == 0)) {
> +	     dp->i_d.di_anextents == 0)))

Also... maybe this should just be a separate predicate?

static inline bool
xfs_attr_is_leaf_format(
	struct xfs_inode	*dp)
{
	return dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
		dp->i_d.di_anextents > 0;
}


if (xfs_attr_is_leaf_format(dp))
	goto add_leaf;

> +		goto sf_to_leaf;
>  
> -		/*
> -		 * Try to add the attr to the attribute list in the inode.
> -		 */
> -		error = xfs_attr_try_sf_addname(dp, args);
> -		if (error != -ENOSPC) {
> -			error2 = xfs_trans_commit(args->trans);
> -			args->trans = NULL;
> -			return error ? error : error2;
> -		}
> +	/*
> +	 * Try to add the attr to the attribute list in the inode.
> +	 */
> +	error = xfs_attr_try_sf_addname(dp, args);
> +	if (error != -ENOSPC) {
> +		error2 = xfs_trans_commit(args->trans);
> +		args->trans = NULL;
> +		return error ? error : error2;
> +	}
>  
> -		/*
> -		 * It won't fit in the shortform, transform to a leaf block.
> -		 * GROT: another possible req'mt for a double-split btree op.
> -		 */
> -		error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
> -		if (error)
> -			return error;
> +	/*
> +	 * It won't fit in the shortform, transform to a leaf block.
> +	 * GROT: another possible req'mt for a double-split btree op.
> +	 */
> +	error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
> +	if (error)
> +		return error;
>  
> -		/*
> -		 * Prevent the leaf buffer from being unlocked so that a
> -		 * concurrent AIL push cannot grab the half-baked leaf
> -		 * buffer and run into problems with the write verifier.
> -		 * Once we're done rolling the transaction we can release
> -		 * the hold and add the attr to the leaf.
> -		 */
> -		xfs_trans_bhold(args->trans, leaf_bp);
> -		error = xfs_defer_finish(&args->trans);
> -		xfs_trans_bhold_release(args->trans, leaf_bp);
> -		if (error) {
> -			xfs_trans_brelse(args->trans, leaf_bp);
> -			return error;
> -		}
> +	/*
> +	 * Prevent the leaf buffer from being unlocked so that a
> +	 * concurrent AIL push cannot grab the half-baked leaf
> +	 * buffer and run into problems with the write verifier.
> +	 * Once we're done rolling the transaction we can release
> +	 * the hold and add the attr to the leaf.
> +	 */
> +	xfs_trans_bhold(args->trans, leaf_bp);
> +	error = xfs_defer_finish(&args->trans);
> +	xfs_trans_bhold_release(args->trans, leaf_bp);
> +	if (error) {
> +		xfs_trans_brelse(args->trans, leaf_bp);
> +		return error;
>  	}
>  
> +sf_to_leaf:

The attr list already has to be in list format by the time it gets here,
which means that the label name is misleading.  How about "add_leaf" ?

--D

>  	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>  		error = xfs_attr_leaf_addname(args);
>  		if (error != -ENOSPC)
> -- 
> 2.7.4
> 

Re: [PATCH v6 15/16] xfs: Add delay ready attr remove routines

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:34PM -0700, Allison Collins wrote:
> This patch modifies the attr remove routines to be delay ready. This
> means they no longer roll or commit transactions, but instead return
> -EAGAIN to have the calling routine roll and refresh the transaction.
> In this series, xfs_attr_remove_args has become xfs_attr_remove_iter,
> which uses a sort of state machine like switch to keep track of where it
> was when EAGAIN was returned. xfs_attr_node_removename has also been
> modified to use the switch, and a  new version of xfs_attr_remove_args
> consists of a simple loop to refresh the transaction until the operation
> is completed.  A helper function xfs_attr_node_shrink has also been
> added to help simplify xfs_attr_node_removename and reduce length.
> 
> This patch also adds a new struct xfs_delattr_context, which we will use
> to keep track of the current state of an attribute operation. The new
> xfs_delattr_state enum is used to track various operations that are in
> progress so that we know not to repeat them, and resume where we left
> off before EAGAIN was returned to cycle out the transaction. Other
> members take the place of local variables that need to retain their
> values across multiple function recalls.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c     | 181 +++++++++++++++++++++++++++++++++----------
>  fs/xfs/libxfs/xfs_attr.h     |   1 +
>  fs/xfs/libxfs/xfs_da_btree.h |  24 ++++++
>  fs/xfs/scrub/common.c        |   2 +
>  fs/xfs/xfs_acl.c             |   2 +
>  fs/xfs/xfs_attr_list.c       |   1 +
>  fs/xfs/xfs_ioctl.c           |   2 +
>  fs/xfs/xfs_ioctl32.c         |   2 +
>  fs/xfs/xfs_iops.c            |   2 +
>  fs/xfs/xfs_xattr.c           |   1 +
>  10 files changed, 176 insertions(+), 42 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index 90e0b2d..92929ad 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -371,11 +371,60 @@ xfs_has_attr(
>   */
>  int
>  xfs_attr_remove_args(
> +	struct xfs_da_args	*args)
> +{
> +	int			error = 0;
> +	int			err2 = 0;
> +
> +	do {
> +		error = xfs_attr_remove_iter(args);
> +		if (error && error != -EAGAIN)
> +			goto out;
> +
> +		if (args->dac.flags & XFS_DAC_FINISH_TRANS) {
> +			args->dac.flags &= ~XFS_DAC_FINISH_TRANS;

Won't rolling the transaction take care of calling defer_finish?

(Granted, I think you started this patch set long before we actually
introduced that behavior...)

> +			err2 = xfs_defer_finish(&args->trans);
> +			if (err2) {
> +				error = err2;
> +				goto out;
> +			}
> +		}
> +
> +		err2 = xfs_trans_roll_inode(&args->trans, args->dp);
> +		if (err2) {
> +			error = err2;
> +			goto out;
> +		}
> +
> +	} while (error == -EAGAIN);

/me suspects this could be compressed down to:

do {
	error = xfs_attr_remove_iter(args);
	if (error == -EAGAIN) {
		error = 0;
		break;
	}
	if (error)
		break;

	error = xfs_trans_roll_inode(...);
} while (error == 0);

return error;

> +out:
> +	return error;
> +}
> +
> +/*
> + * Remove the attribute specified in @args.
> + * This routine is meant to function as a delayed operation, and may return
> + * -EGAIN when the transaction needs to be rolled.  Calling functions will need
> + * to handle this, and recall the function until a successful error code is
> + * returned.

"This function may return -EAGAIN to signal that the transaction needs
to be rolled.  Callers should continue calling this function until they
receive a return value other than -EAGAIN." ?

> + */
> +int
> +xfs_attr_remove_iter(
>  	struct xfs_da_args      *args)
>  {
>  	struct xfs_inode	*dp = args->dp;
>  	int			error;
>  
> +	/* State machine switch */
> +	switch (args->dac.dela_state) {
> +	case XFS_DAS_RM_SHRINK:
> +	case XFS_DAS_RM_NODE_BLKS:
> +		goto node;
> +	default:
> +		break;
> +	}
> +
>  	if (!xfs_inode_hasattr(dp)) {
>  		error = -ENOATTR;
>  	} else if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
> @@ -384,6 +433,7 @@ xfs_attr_remove_args(
>  	} else if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>  		error = xfs_attr_leaf_removename(args);
>  	} else {
> +node:
>  		error = xfs_attr_node_removename(args);
>  	}
>  
> @@ -887,9 +937,8 @@ xfs_attr_leaf_removename(
>  		/* bp is gone due to xfs_da_shrink_inode */
>  		if (error)
>  			return error;
> -		error = xfs_defer_finish(&args->trans);
> -		if (error)
> -			return error;
> +
> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>  	}
>  	return 0;
>  }
> @@ -1233,6 +1282,42 @@ xfs_attr_init_unmapstate(
>  	return 0;
>  }
>  
> +/*
> + * Shrink an attribute from leaf to shortform
> + */
> +STATIC int
> +xfs_attr_node_shrink(
> +	struct xfs_da_args	*args,
> +	struct xfs_da_state     *state)
> +{
> +	struct xfs_inode	*dp = args->dp;
> +	int			error, forkoff;
> +	struct xfs_buf		*bp;
> +
> +	/*
> +	 * Have to get rid of the copy of this dabuf in the state.
> +	 */
> +	ASSERT(state->path.active == 1);
> +	ASSERT(state->path.blk[0].bp);
> +	state->path.blk[0].bp = NULL;
> +
> +	error = xfs_attr3_leaf_read(args->trans, args->dp, 0, &bp);
> +	if (error)
> +		return error;
> +
> +	forkoff = xfs_attr_shortform_allfit(bp, dp);
> +	if (forkoff) {
> +		error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
> +		/* bp is gone due to xfs_da_shrink_inode */
> +		if (error)
> +			return error;
> +
> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
> +	} else
> +		xfs_trans_brelse(args->trans, bp);
> +
> +	return 0;
> +}
>  
>  /*
>   * Remove a name from a B-tree attribute list.
> @@ -1240,6 +1325,11 @@ xfs_attr_init_unmapstate(
>   * This will involve walking down the Btree, and may involve joining
>   * leaf nodes and even joining intermediate nodes up to and including
>   * the root node (a special case of an intermediate node).
> + *
> + * This routine is meant to function as either an inline or delayed operation,
> + * and may return -EAGAIN when the transaction needs to be rolled.  Calling
> + * functions will need to handle this, and recall the function until a
> + * successful error code is returned.
>   */
>  STATIC int
>  xfs_attr_node_removename(
> @@ -1247,15 +1337,28 @@ xfs_attr_node_removename(
>  {
>  	struct xfs_da_state	*state;
>  	struct xfs_da_state_blk	*blk;
> -	struct xfs_buf		*bp;
> -	int			retval, error, forkoff;
> +	int			retval, error;
>  	struct xfs_inode	*dp = args->dp;
>  
>  	trace_xfs_attr_node_removename(args);
> +	state = args->dac.da_state;
> +	blk = args->dac.blk;
> +
> +	/* State machine switch */
> +	switch (args->dac.dela_state) {
> +	case XFS_DAS_RM_NODE_BLKS:
> +		goto rm_node_blks;
> +	case XFS_DAS_RM_SHRINK:
> +		goto rm_shrink;
> +	default:
> +		break;
> +	}
>  
>  	error = xfs_attr_node_hasname(args, &state);
>  	if (error != -EEXIST)
>  		goto out;
> +	else
> +		error = 0;
>  
>  	/*
>  	 * If there is an out-of-line value, de-allocate the blocks.
> @@ -1265,18 +1368,36 @@ xfs_attr_node_removename(
>  	blk = &state->path.blk[ state->path.active-1 ];
>  	ASSERT(blk->bp != NULL);
>  	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
> +
> +	/*
> +	 * Store blk and state in the context incase we need to cycle out the
> +	 * transaction
> +	 */
> +	args->dac.blk = blk;
> +	args->dac.da_state = state;
> +
>  	if (args->rmtblkno > 0) {
>  		error = xfs_attr_init_unmapstate(args, state);
>  		if (error)
>  			goto out;
>  
> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
> +		error = xfs_attr_rmtval_invalidate(args);
>  		if (error)
>  			goto out;
> +	}
>  
> -		error = xfs_attr_rmtval_remove(args);
> -		if (error)
> -			goto out;
> +rm_node_blks:
> +
> +	args->dac.dela_state = XFS_DAS_RM_NODE_BLKS;

That's odd -- we jumped here if dela_state == RM_NODE_BLKS, so why do we
need to set dela_state?  Does that put us at risk of jumping back here
by accident?

Oh, maybe this is supposed to make sure we keep coming back for more
bunmapi if the the remote value blocks were backed by multiple extents.

Perhaps this would be better named DAS_RMTVAL_REMOVE?

> +
> +	if (args->rmtblkno > 0) {
> +		error = xfs_attr_rmtval_unmap(args);
> +
> +		if (error) {
> +			if (error == -EAGAIN)
> +				args->dac.dela_state = XFS_DAS_RM_NODE_BLKS;

(Didn't we already set this state?)

> +			return error;
> +		}
>  
>  		/*
>  		 * Refill the state structure with buffers, the prior calls
> @@ -1302,44 +1423,20 @@ xfs_attr_node_removename(
>  		error = xfs_da3_join(state);
>  		if (error)
>  			goto out;
> -		error = xfs_defer_finish(&args->trans);
> -		if (error)
> -			goto out;
> -		/*
> -		 * Commit the Btree join operation and start a new trans.
> -		 */
> -		error = xfs_trans_roll_inode(&args->trans, dp);
> -		if (error)
> -			goto out;
> +
> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
> +		args->dac.dela_state = XFS_DAS_RM_SHRINK;
> +		return -EAGAIN;
>  	}
>  
> +rm_shrink:
> +	args->dac.dela_state = XFS_DAS_RM_SHRINK;
> +
>  	/*
>  	 * If the result is small enough, push it all into the inode.
>  	 */
> -	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
> -		/*
> -		 * Have to get rid of the copy of this dabuf in the state.
> -		 */
> -		ASSERT(state->path.active == 1);
> -		ASSERT(state->path.blk[0].bp);
> -		state->path.blk[0].bp = NULL;
> -
> -		error = xfs_attr3_leaf_read(args->trans, args->dp, 0, &bp);
> -		if (error)
> -			goto out;
> -
> -		if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
> -			error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
> -			/* bp is gone due to xfs_da_shrink_inode */
> -			if (error)
> -				goto out;
> -			error = xfs_defer_finish(&args->trans);
> -			if (error)
> -				goto out;
> -		} else
> -			xfs_trans_brelse(args->trans, bp);
> -	}
> -	error = 0;
> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK))
> +		error = xfs_attr_node_shrink(args, state);
>  
>  out:
>  	if (state)
> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
> index ce7b039..ea873a5 100644
> --- a/fs/xfs/libxfs/xfs_attr.h
> +++ b/fs/xfs/libxfs/xfs_attr.h
> @@ -155,6 +155,7 @@ int xfs_attr_set_args(struct xfs_da_args *args);
>  int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
>  int xfs_has_attr(struct xfs_da_args *args);
>  int xfs_attr_remove_args(struct xfs_da_args *args);
> +int xfs_attr_remove_iter(struct xfs_da_args *args);
>  int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
>  		  int flags, struct attrlist_cursor_kern *cursor);
>  bool xfs_attr_namecheck(const void *name, size_t length);
> diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
> index 14f1be3..7fc87da 100644
> --- a/fs/xfs/libxfs/xfs_da_btree.h
> +++ b/fs/xfs/libxfs/xfs_da_btree.h
> @@ -50,9 +50,33 @@ enum xfs_dacmp {
>  };
>  
>  /*
> + * Enum values for xfs_delattr_context.da_state
> + */

I can tell this is an enum. ;)

Could you have this comment explain a little more about how states are
managed?  I /think/ these state values mark where we were in the attr
code when something decided it was time to roll a transaction, right?

In other words, dela_state is a save point where we jump back to after
hopping out to roll a transaction?

(The parts of the code we jump somewhere based on dela_state and
immediately set dela_state confused me, so I don't know if I've
understood how the state machine works correctly.)

> +enum xfs_delattr_state {
> +	XFS_DAS_RM_SHRINK	= 1, /* We are shrinking the tree */

This means that dela_state == 0 means "do not use delayed attributes",
right?  We might want to make the meaning of zero more explicit.

> +	XFS_DAS_RM_NODE_BLKS	= 2, /* We are removing node blocks */

Also, the C compiler will auto-increment this for you if you don't
assign an explicit value.

> +};
> +
> +/*
> + * Defines for xfs_delattr_context.flags
> + */
> +#define	XFS_DAC_FINISH_TRANS	0x1 /* indicates to finish the transaction */
> +
> +/*
> + * Context used for keeping track of delayed attribute operations
> + */
> +struct xfs_delattr_context {
> +	struct xfs_da_state	*da_state;
> +	struct xfs_da_state_blk *blk;
> +	int			flags;

unsigned int, assuming FINISH_TRANS doesn't just disappear.

> +	enum xfs_delattr_state	dela_state;
> +};
> +
> +/*
>   * Structure to ease passing around component names.
>   */
>  typedef struct xfs_da_args {
> +	struct xfs_delattr_context dac; /* context used for delay attr ops */
>  	struct xfs_da_geometry *geo;	/* da block geometry */
>  	struct xfs_name	name;		/* name, length and argument  flags*/
>  	uint8_t		filetype;	/* filetype of inode for directories */
> diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
> index 1887605..9a649d1 100644
> --- a/fs/xfs/scrub/common.c
> +++ b/fs/xfs/scrub/common.c
> @@ -24,6 +24,8 @@
>  #include "xfs_rmap_btree.h"
>  #include "xfs_log.h"
>  #include "xfs_trans_priv.h"
> +#include "xfs_da_format.h"
> +#include "xfs_da_btree.h"

Are these necessary if all the da_args stuff moves to xfs_types.h?

--D

>  #include "xfs_attr.h"
>  #include "xfs_reflink.h"
>  #include "scrub/scrub.h"
> diff --git a/fs/xfs/xfs_acl.c b/fs/xfs/xfs_acl.c
> index 42ac847..d65e6d8 100644
> --- a/fs/xfs/xfs_acl.c
> +++ b/fs/xfs/xfs_acl.c
> @@ -10,6 +10,8 @@
>  #include "xfs_trans_resv.h"
>  #include "xfs_mount.h"
>  #include "xfs_inode.h"
> +#include "xfs_da_format.h"
> +#include "xfs_da_btree.h"
>  #include "xfs_attr.h"
>  #include "xfs_trace.h"
>  #include "xfs_error.h"
> diff --git a/fs/xfs/xfs_attr_list.c b/fs/xfs/xfs_attr_list.c
> index d37743b..881b9a4 100644
> --- a/fs/xfs/xfs_attr_list.c
> +++ b/fs/xfs/xfs_attr_list.c
> @@ -12,6 +12,7 @@
>  #include "xfs_trans_resv.h"
>  #include "xfs_mount.h"
>  #include "xfs_da_format.h"
> +#include "xfs_da_btree.h"
>  #include "xfs_inode.h"
>  #include "xfs_trans.h"
>  #include "xfs_bmap.h"
> diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> index 28c07c9..7c1d9da 100644
> --- a/fs/xfs/xfs_ioctl.c
> +++ b/fs/xfs/xfs_ioctl.c
> @@ -15,6 +15,8 @@
>  #include "xfs_iwalk.h"
>  #include "xfs_itable.h"
>  #include "xfs_error.h"
> +#include "xfs_da_format.h"
> +#include "xfs_da_btree.h"
>  #include "xfs_attr.h"
>  #include "xfs_bmap.h"
>  #include "xfs_bmap_util.h"
> diff --git a/fs/xfs/xfs_ioctl32.c b/fs/xfs/xfs_ioctl32.c
> index 769581a..d504f8f 100644
> --- a/fs/xfs/xfs_ioctl32.c
> +++ b/fs/xfs/xfs_ioctl32.c
> @@ -17,6 +17,8 @@
>  #include "xfs_itable.h"
>  #include "xfs_fsops.h"
>  #include "xfs_rtalloc.h"
> +#include "xfs_da_format.h"
> +#include "xfs_da_btree.h"
>  #include "xfs_attr.h"
>  #include "xfs_ioctl.h"
>  #include "xfs_ioctl32.h"
> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> index e85bbf5..a2d299f 100644
> --- a/fs/xfs/xfs_iops.c
> +++ b/fs/xfs/xfs_iops.c
> @@ -13,6 +13,8 @@
>  #include "xfs_inode.h"
>  #include "xfs_acl.h"
>  #include "xfs_quota.h"
> +#include "xfs_da_format.h"
> +#include "xfs_da_btree.h"
>  #include "xfs_attr.h"
>  #include "xfs_trans.h"
>  #include "xfs_trace.h"
> diff --git a/fs/xfs/xfs_xattr.c b/fs/xfs/xfs_xattr.c
> index 74133a5..d8dc72d 100644
> --- a/fs/xfs/xfs_xattr.c
> +++ b/fs/xfs/xfs_xattr.c
> @@ -10,6 +10,7 @@
>  #include "xfs_log_format.h"
>  #include "xfs_da_format.h"
>  #include "xfs_inode.h"
> +#include "xfs_da_btree.h"
>  #include "xfs_attr.h"
>  #include "xfs_acl.h"
>  
> -- 
> 2.7.4
> 

Re: [PATCH v6 16/16] xfs: Add delay ready attr set routines

[Darrick J. Wong]

On Sat, Jan 18, 2020 at 03:50:35PM -0700, Allison Collins wrote:
> This patch modifies the attr set routines to be delay ready. This means
> they no longer roll or commit transactions, but instead return -EAGAIN
> to have the calling routine roll and refresh the transaction.  In this
> series, xfs_attr_set_args has become xfs_attr_set_iter, which uses a
> state machine like switch to keep track of where it was when EAGAIN was
> returned.
> 
> Part of xfs_attr_leaf_addname has been factored out into a new helper
> function xfs_attr_leaf_try_add to allow transaction cycling between the
> two routines.
> 
> Two new helper functions have been added: xfs_attr_rmtval_set_init and
> xfs_attr_rmtval_set_blk.  They provide a subset of logic similar to
> xfs_attr_rmtval_set, but they store the current block in the delay attr
> context to allow the caller to roll the transaction between allocations.
> This helps to simplify and consolidate code used by
> xfs_attr_leaf_addname and xfs_attr_node_addname
> 
> Finally, xfs_attr_set_args has become a simple loop to refresh the
> transaction until the operation is completed.
> 
> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_attr.c        | 347 +++++++++++++++++++++++++---------------
>  fs/xfs/libxfs/xfs_attr.h        |   1 +
>  fs/xfs/libxfs/xfs_attr_remote.c |  67 +++++++-
>  fs/xfs/libxfs/xfs_attr_remote.h |   4 +
>  fs/xfs/libxfs/xfs_da_btree.h    |  13 ++
>  5 files changed, 301 insertions(+), 131 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> index 92929ad..02a7f55 100644
> --- a/fs/xfs/libxfs/xfs_attr.c
> +++ b/fs/xfs/libxfs/xfs_attr.c
> @@ -58,6 +58,7 @@ STATIC int xfs_attr_node_hasname(xfs_da_args_t *args,
>  				 struct xfs_da_state **state);
>  STATIC int xfs_attr_fillstate(xfs_da_state_t *state);
>  STATIC int xfs_attr_refillstate(xfs_da_state_t *state);
> +STATIC int xfs_attr_leaf_try_add(struct xfs_da_args *args, struct xfs_buf *bp);
>  
>  
>  STATIC int
> @@ -259,9 +260,86 @@ int
>  xfs_attr_set_args(
>  	struct xfs_da_args	*args)
>  {
> +	int			error = 0;
> +	int			err2 = 0;
> +	struct xfs_buf		*leaf_bp = NULL;
> +
> +	do {
> +		error = xfs_attr_set_iter(args, &leaf_bp);
> +		if (error && error != -EAGAIN)
> +			goto out;
> +
> +		if (args->dac.flags & XFS_DAC_FINISH_TRANS) {
> +			args->dac.flags &= ~XFS_DAC_FINISH_TRANS;
> +
> +			err2 = xfs_defer_finish(&args->trans);
> +			if (err2) {
> +				error = err2;
> +				goto out;
> +			}
> +		}
> +
> +		err2 = xfs_trans_roll_inode(&args->trans, args->dp);
> +		if (err2) {
> +			error = err2;
> +			goto out;
> +		}
> +
> +		if (leaf_bp) {
> +			xfs_trans_bjoin(args->trans, leaf_bp);
> +			xfs_trans_bhold(args->trans, leaf_bp);
> +		}
> +
> +	} while (error == -EAGAIN);
> +
> +out:
> +	return error;
> +}
> +
> +/*
> + * Set the attribute specified in @args.
> + * This routine is meant to function as a delayed operation, and may return
> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
> + * to handle this, and recall the function until a successful error code is
> + * returned.
> + */
> +int
> +xfs_attr_set_iter(
> +	struct xfs_da_args	*args,
> +	struct xfs_buf          **leaf_bp)
> +{
>  	struct xfs_inode	*dp = args->dp;
> -	struct xfs_buf          *leaf_bp = NULL;
> -	int			error, error2 = 0;
> +	int			error = 0;
> +	int			sf_size;
> +
> +	/* State machine switch */
> +	switch (args->dac.dela_state) {
> +	case XFS_DAS_SF_TO_LEAF:
> +		goto sf_to_leaf;
> +	case XFS_DAS_ALLOC_LEAF:
> +	case XFS_DAS_FLIP_LFLAG:
> +	case XFS_DAS_FOUND_LBLK:
> +		goto leaf;
> +	case XFS_DAS_FOUND_NBLK:
> +	case XFS_DAS_FLIP_NFLAG:
> +	case XFS_DAS_ALLOC_NODE:
> +	case XFS_DAS_LEAF_TO_NODE:
> +		goto node;
> +	default:

Er... do these need to ASSERT on invalid states?  e.g. if somehow we end
up with XFS_DAS_RM_SHRINK, we ought to blow up loudly, right?


> +		break;
> +	}
> +
> +	/*
> +	 * New inodes may not have an attribute fork yet. So set the attribute
> +	 * fork appropriately
> +	 */
> +	if (XFS_IFORK_Q((args->dp)) == 0) {
> +		sf_size = sizeof(struct xfs_attr_sf_hdr) +
> +		     XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
> +		xfs_bmap_set_attrforkoff(args->dp, sf_size, NULL);
> +		args->dp->i_afp = kmem_zone_zalloc(xfs_ifork_zone, 0);
> +		args->dp->i_afp->if_flags = XFS_IFEXTENTS;
> +	}
>  
>  	/*
>  	 * If the attribute list is non-existent or a shortform list, proceed to
> @@ -272,22 +350,20 @@ xfs_attr_set_args(
>  	    (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
>  	     dp->i_d.di_anextents == 0)))
>  		goto sf_to_leaf;
> -
>  	/*
>  	 * Try to add the attr to the attribute list in the inode.
>  	 */
>  	error = xfs_attr_try_sf_addname(dp, args);
> -	if (error != -ENOSPC) {
> -		error2 = xfs_trans_commit(args->trans);
> -		args->trans = NULL;
> -		return error ? error : error2;

Where does this commit go?

> -	}
> +
> +	/* Should only be 0, -EEXIST or ENOSPC */
> +	if (error != -ENOSPC)
> +		return error;
>  
>  	/*
>  	 * It won't fit in the shortform, transform to a leaf block.
>  	 * GROT: another possible req'mt for a double-split btree op.
>  	 */
> -	error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
> +	error = xfs_attr_shortform_to_leaf(args, leaf_bp);
>  	if (error)
>  		return error;
>  
> @@ -295,42 +371,47 @@ xfs_attr_set_args(
>  	 * Prevent the leaf buffer from being unlocked so that a
>  	 * concurrent AIL push cannot grab the half-baked leaf
>  	 * buffer and run into problems with the write verifier.
> -	 * Once we're done rolling the transaction we can release
> -	 * the hold and add the attr to the leaf.
>  	 */
> -	xfs_trans_bhold(args->trans, leaf_bp);
> -	error = xfs_defer_finish(&args->trans);
> -	xfs_trans_bhold_release(args->trans, leaf_bp);
> -	if (error) {
> -		xfs_trans_brelse(args->trans, leaf_bp);
> -		return error;
> -	}
> +	xfs_trans_bhold(args->trans, *leaf_bp);
> +	args->dac.flags |= XFS_DAC_FINISH_TRANS;
> +	args->dac.dela_state = XFS_DAS_SF_TO_LEAF;
> +	return -EAGAIN;
>  
>  sf_to_leaf:
>  
> -	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
> -		error = xfs_attr_leaf_addname(args);
> -		if (error != -ENOSPC)
> -			return error;
> -
> -		/*
> -		 * Commit that transaction so that the node_addname()
> -		 * call can manage its own transactions.
> -		 */
> -		error = xfs_defer_finish(&args->trans);
> -		if (error)
> -			return error;
> +	/*
> +	 * After a shortform to leaf conversion, we need to hold the leaf and
> +	 * cylce out the transaction.  When we get back, we need to release
> +	 * the leaf.
> +	 */
> +	if (*leaf_bp != NULL) {
> +		xfs_trans_brelse(args->trans, *leaf_bp);
> +		*leaf_bp = NULL;
> +	}
>  
> -		/*
> -		 * Commit the current trans (including the inode) and
> -		 * start a new one.
> -		 */
> -		error = xfs_trans_roll_inode(&args->trans, dp);
> -		if (error)
> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
> +		error = xfs_attr_leaf_try_add(args, *leaf_bp);
> +		switch (error) {
> +		case -ENOSPC:
> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
> +			args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
> +			return -EAGAIN;
> +		case 0:
> +			args->dac.dela_state = XFS_DAS_FOUND_LBLK;
> +			return -EAGAIN;
> +		default:
>  			return error;
> -
> +		}
> +leaf:
> +		error = xfs_attr_leaf_addname(args);
> +		if (error == -ENOSPC) {
> +			args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
> +			return -EAGAIN;
> +		}
> +		return error;
>  	}
> -
> +	args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
> +node:
>  	error = xfs_attr_node_addname(args);
>  	return error;
>  }
> @@ -759,27 +840,28 @@ xfs_attr_leaf_try_add(
>   *
>   * This leaf block cannot have a "remote" value, we only call this routine
>   * if bmap_one_block() says there is only one block (ie: no remote blks).
> + *
> + * This routine is meant to function as a delayed operation, and may return
> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
> + * to handle this, and recall the function until a successful error code is
> + * returned.
>   */
>  STATIC int
>  xfs_attr_leaf_addname(struct xfs_da_args	*args)
>  {
> -	int			error, forkoff;
>  	struct xfs_buf		*bp = NULL;
> +	int			error, forkoff;
>  	struct xfs_inode	*dp = args->dp;
>  
> -	trace_xfs_attr_leaf_addname(args);
> -
> -	error = xfs_attr_leaf_try_add(args, bp);
> -	if (error)
> -		return error;
> -
> -	/*
> -	 * Commit the transaction that added the attr name so that
> -	 * later routines can manage their own transactions.
> -	 */
> -	error = xfs_trans_roll_inode(&args->trans, dp);
> -	if (error)
> -		return error;
> +	/* State machine switch */
> +	switch (args->dac.dela_state) {
> +	case XFS_DAS_FLIP_LFLAG:
> +		goto flip_flag;
> +	case XFS_DAS_ALLOC_LEAF:
> +		goto alloc_leaf;
> +	default:
> +		break;
> +	}
>  
>  	/*
>  	 * If there was an out-of-line value, allocate the blocks we
> @@ -788,7 +870,28 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>  	 * maximum size of a transaction and/or hit a deadlock.
>  	 */
>  	if (args->rmtblkno > 0) {
> -		error = xfs_attr_rmtval_set(args);
> +
> +		/* Open coded xfs_attr_rmtval_set without trans handling */
> +		error = xfs_attr_rmtval_set_init(args);
> +		if (error)
> +			return error;
> +
> +		/*
> +		 * Roll through the "value", allocating blocks on disk as
> +		 * required.
> +		 */
> +alloc_leaf:
> +		while (args->dac.blkcnt > 0) {
> +			error = xfs_attr_rmtval_set_blk(args);
> +			if (error)
> +				return error;
> +
> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
> +			args->dac.dela_state = XFS_DAS_ALLOC_LEAF;
> +			return -EAGAIN;
> +		}
> +
> +		error = xfs_attr_rmtval_set_value(args);
>  		if (error)
>  			return error;
>  	}
> @@ -807,13 +910,10 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>  		error = xfs_attr3_leaf_flipflags(args);
>  		if (error)
>  			return error;
> -		/*
> -		 * Commit the flag value change and start the next trans in
> -		 * series.
> -		 */
> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
> -		if (error)
> -			return error;
> +
> +		args->dac.dela_state = XFS_DAS_FLIP_LFLAG;
> +		return -EAGAIN;
> +flip_flag:
>  
>  		/*
>  		 * Dismantle the "old" attribute/value pair by removing
> @@ -844,34 +944,17 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>  		/*
>  		 * If the result is small enough, shrink it all into the inode.
>  		 */
> -		if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
> +		forkoff = xfs_attr_shortform_allfit(bp, dp);
> +		if (forkoff)
>  			error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
> -			/* bp is gone due to xfs_da_shrink_inode */
> -			if (error)
> -				return error;
> -			error = xfs_defer_finish(&args->trans);
> -			if (error)
> -				return error;
> -		}
>  
> -		/*
> -		 * Commit the remove and start the next trans in series.
> -		 */
> -		error = xfs_trans_roll_inode(&args->trans, dp);
> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>  
>  	} else if (args->rmtblkno > 0) {
>  		/*
>  		 * Added a "remote" value, just clear the incomplete flag.
>  		 */
>  		error = xfs_attr3_leaf_clearflag(args);
> -		if (error)
> -			return error;
> -
> -		/*
> -		 * Commit the flag value change and start the next trans in
> -		 * series.
> -		 */
> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>  	}
>  	return error;
>  }
> @@ -1020,16 +1103,22 @@ xfs_attr_node_hasname(
>   *
>   * "Remote" attribute values confuse the issue and atomic rename operations
>   * add a whole extra layer of confusion on top of that.
> + *
> + * This routine is meant to function as a delayed operation, and may return
> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
> + * to handle this, and recall the function until a successful error code is
> + *returned.
>   */
>  STATIC int
>  xfs_attr_node_addname(
>  	struct xfs_da_args	*args)
>  {
> -	struct xfs_da_state	*state;
> +	struct xfs_da_state	*state = NULL;
>  	struct xfs_da_state_blk	*blk;
>  	struct xfs_inode	*dp;
>  	struct xfs_mount	*mp;
> -	int			retval, error;
> +	int			retval = 0;
> +	int			error = 0;
>  
>  	trace_xfs_attr_node_addname(args);
>  
> @@ -1038,7 +1127,19 @@ xfs_attr_node_addname(
>  	 */
>  	dp = args->dp;
>  	mp = dp->i_mount;
> -restart:
> +
> +	/* State machine switch */
> +	switch (args->dac.dela_state) {
> +	case XFS_DAS_FLIP_NFLAG:
> +		goto flip_flag;
> +	case XFS_DAS_FOUND_NBLK:
> +		goto found_nblk;
> +	case XFS_DAS_ALLOC_NODE:
> +		goto alloc_node;
> +	default:
> +		break;
> +	}
> +
>  	/*
>  	 * Search to see if name already exists, and get back a pointer
>  	 * to where it should go.
> @@ -1088,19 +1189,13 @@ xfs_attr_node_addname(
>  			error = xfs_attr3_leaf_to_node(args);
>  			if (error)
>  				goto out;
> -			error = xfs_defer_finish(&args->trans);
> -			if (error)
> -				goto out;
>  
>  			/*
> -			 * Commit the node conversion and start the next
> -			 * trans in the chain.
> +			 * Restart routine from the top.  No need to set  the
> +			 * state

*OH*, dela_state == 0 does not mean "do not use delayed attrs", we're
switching /all/ users to delayed attrs.  Urk.

>  			 */
> -			error = xfs_trans_roll_inode(&args->trans, dp);
> -			if (error)
> -				goto out;
> -
> -			goto restart;
> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
> +			return -EAGAIN;
>  		}
>  
>  		/*
> @@ -1112,9 +1207,7 @@ xfs_attr_node_addname(
>  		error = xfs_da3_split(state);
>  		if (error)
>  			goto out;
> -		error = xfs_defer_finish(&args->trans);
> -		if (error)
> -			goto out;
> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>  	} else {
>  		/*
>  		 * Addition succeeded, update Btree hashvals.
> @@ -1129,13 +1222,9 @@ xfs_attr_node_addname(
>  	xfs_da_state_free(state);
>  	state = NULL;
>  
> -	/*
> -	 * Commit the leaf addition or btree split and start the next
> -	 * trans in the chain.
> -	 */
> -	error = xfs_trans_roll_inode(&args->trans, dp);
> -	if (error)
> -		goto out;
> +	args->dac.dela_state = XFS_DAS_FOUND_NBLK;
> +	return -EAGAIN;
> +found_nblk:
>  
>  	/*
>  	 * If there was an out-of-line value, allocate the blocks we
> @@ -1144,7 +1233,27 @@ xfs_attr_node_addname(
>  	 * maximum size of a transaction and/or hit a deadlock.
>  	 */
>  	if (args->rmtblkno > 0) {
> -		error = xfs_attr_rmtval_set(args);
> +		/* Open coded xfs_attr_rmtval_set without trans handling */
> +		error = xfs_attr_rmtval_set_init(args);
> +		if (error)
> +			return error;
> +
> +		/*
> +		 * Roll through the "value", allocating blocks on disk as
> +		 * required.
> +		 */
> +alloc_node:
> +		while (args->dac.blkcnt > 0) {
> +			error = xfs_attr_rmtval_set_blk(args);
> +			if (error)
> +				return error;
> +
> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
> +			args->dac.dela_state = XFS_DAS_ALLOC_NODE;
> +			return -EAGAIN;
> +		}
> +
> +		error = xfs_attr_rmtval_set_value(args);
>  		if (error)
>  			return error;
>  	}
> @@ -1167,10 +1276,9 @@ xfs_attr_node_addname(
>  		 * Commit the flag value change and start the next trans in
>  		 * series
>  		 */
> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
> -		if (error)
> -			goto out;
> -
> +		args->dac.dela_state = XFS_DAS_FLIP_NFLAG;
> +		return -EAGAIN;
> +flip_flag:
>  		/*
>  		 * Dismantle the "old" attribute/value pair by removing
>  		 * a "remote" value (if it exists).
> @@ -1199,7 +1307,6 @@ xfs_attr_node_addname(
>  		error = xfs_da3_node_lookup_int(state, &retval);
>  		if (error)
>  			goto out;
> -
>  		/*
>  		 * Remove the name and update the hashvals in the tree.
>  		 */
> @@ -1207,7 +1314,6 @@ xfs_attr_node_addname(
>  		ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
>  		error = xfs_attr3_leaf_remove(blk->bp, args);
>  		xfs_da3_fixhashpath(state, &state->path);
> -
>  		/*
>  		 * Check to see if the tree needs to be collapsed.
>  		 */
> @@ -1215,18 +1321,9 @@ xfs_attr_node_addname(
>  			error = xfs_da3_join(state);
>  			if (error)
>  				goto out;
> -			error = xfs_defer_finish(&args->trans);
> -			if (error)
> -				goto out;
> -		}
> -
> -		/*
> -		 * Commit and start the next trans in the chain.
> -		 */
> -		error = xfs_trans_roll_inode(&args->trans, dp);
> -		if (error)
> -			goto out;
>  
> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
> +		}
>  	} else if (args->rmtblkno > 0) {
>  		/*
>  		 * Added a "remote" value, just clear the incomplete flag.
> @@ -1234,14 +1331,6 @@ xfs_attr_node_addname(
>  		error = xfs_attr3_leaf_clearflag(args);
>  		if (error)
>  			goto out;
> -
> -		 /*
> -		  * Commit the flag value change and start the next trans in
> -		  * series.
> -		  */
> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
> -		if (error)
> -			goto out;
>  	}
>  	retval = error = 0;
>  
> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
> index ea873a5..f450d8c 100644
> --- a/fs/xfs/libxfs/xfs_attr.h
> +++ b/fs/xfs/libxfs/xfs_attr.h
> @@ -152,6 +152,7 @@ int xfs_attr_get(struct xfs_inode *ip, struct xfs_name *name,
>  int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
>  		 unsigned char *value, int valuelen, int flags);
>  int xfs_attr_set_args(struct xfs_da_args *args);
> +int xfs_attr_set_iter(struct xfs_da_args *args, struct xfs_buf **leaf_bp);
>  int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
>  int xfs_has_attr(struct xfs_da_args *args);
>  int xfs_attr_remove_args(struct xfs_da_args *args);
> diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
> index f2ee0b8..ad4f14a 100644
> --- a/fs/xfs/libxfs/xfs_attr_remote.c
> +++ b/fs/xfs/libxfs/xfs_attr_remote.c
> @@ -428,7 +428,7 @@ xfs_attr_rmtval_get(
>   * Find a "hole" in the attribute address space large enough for us to drop the
>   * new attribute's value into
>   */
> -STATIC int
> +int
>  xfs_attr_rmt_find_hole(
>  	struct xfs_da_args	*args)
>  {
> @@ -455,7 +455,7 @@ xfs_attr_rmt_find_hole(
>  	return 0;
>  }
>  
> -STATIC int
> +int
>  xfs_attr_rmtval_set_value(
>  	struct xfs_da_args	*args)
>  {
> @@ -588,6 +588,69 @@ xfs_attr_rmtval_set(
>  }
>  
>  /*
> + * Find a hole for the attr and store it in the delayed attr context.  This
> + * initializes the context to roll through allocating an attr extent for a
> + * delayed attr operation
> + */
> +int
> +xfs_attr_rmtval_set_init(
> +	struct xfs_da_args	*args)
> +{
> +	struct xfs_bmbt_irec	*map = &args->dac.map;
> +	int error;
> +
> +	args->dac.lblkno = 0;
> +	args->dac.lfileoff = 0;
> +	args->dac.blkcnt = 0;
> +	args->rmtblkcnt = 0;
> +	args->rmtblkno = 0;
> +	memset(map, 0, sizeof(struct xfs_bmbt_irec));
> +
> +	error = xfs_attr_rmt_find_hole(args);
> +	if (error)
> +		return error;
> +
> +	args->dac.blkcnt = args->rmtblkcnt;
> +	args->dac.lblkno = args->rmtblkno;
> +
> +	return error;
> +}
> +
> +/*
> + * Write one block of the value associated with an attribute into the
> + * out-of-line buffer that we have defined for it. This is similar to a subset
> + * of xfs_attr_rmtval_set, but records the current block to the delayed attr
> + * context, and leaves transaction handling to the caller.
> + */
> +int
> +xfs_attr_rmtval_set_blk(
> +	struct xfs_da_args	*args)
> +{
> +	struct xfs_inode	*dp = args->dp;
> +	struct xfs_bmbt_irec	*map = &args->dac.map;
> +	int nmap;
> +	int error;
> +
> +	nmap = 1;
> +	error = xfs_bmapi_write(args->trans, dp,
> +		  (xfs_fileoff_t)args->dac.lblkno,
> +		  args->dac.blkcnt, XFS_BMAPI_ATTRFORK,
> +		  args->total, map, &nmap);
> +	if (error)
> +		return error;
> +
> +	ASSERT(nmap == 1);
> +	ASSERT((map->br_startblock != DELAYSTARTBLOCK) &&
> +	       (map->br_startblock != HOLESTARTBLOCK));
> +
> +	/* roll attribute extent map forwards */
> +	args->dac.lblkno += map->br_blockcount;
> +	args->dac.blkcnt -= map->br_blockcount;
> +
> +	return 0;
> +}
> +
> +/*
>   * Remove the value associated with an attribute by deleting the
>   * out-of-line buffer that it is stored on.
>   */
> diff --git a/fs/xfs/libxfs/xfs_attr_remote.h b/fs/xfs/libxfs/xfs_attr_remote.h
> index 7ab3770..ab03519 100644
> --- a/fs/xfs/libxfs/xfs_attr_remote.h
> +++ b/fs/xfs/libxfs/xfs_attr_remote.h
> @@ -13,4 +13,8 @@ int xfs_attr_rmtval_set(struct xfs_da_args *args);
>  int xfs_attr_rmtval_remove(struct xfs_da_args *args);
>  int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
>  int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
> +int xfs_attr_rmt_find_hole(struct xfs_da_args *args);
> +int xfs_attr_rmtval_set_value(struct xfs_da_args *args);
> +int xfs_attr_rmtval_set_blk(struct xfs_da_args *args);
> +int xfs_attr_rmtval_set_init(struct xfs_da_args *args);
>  #endif /* __XFS_ATTR_REMOTE_H__ */
> diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
> index 7fc87da..9943062 100644
> --- a/fs/xfs/libxfs/xfs_da_btree.h
> +++ b/fs/xfs/libxfs/xfs_da_btree.h
> @@ -55,6 +55,14 @@ enum xfs_dacmp {
>  enum xfs_delattr_state {
>  	XFS_DAS_RM_SHRINK	= 1, /* We are shrinking the tree */
>  	XFS_DAS_RM_NODE_BLKS	= 2, /* We are removing node blocks */
> +	XFS_DAS_SF_TO_LEAF	= 3, /* Converted short form to leaf */
> +	XFS_DAS_FOUND_LBLK	= 4, /* We found leaf blk for attr */
> +	XFS_DAS_LEAF_TO_NODE	= 5, /* Converted leaf to node */
> +	XFS_DAS_FOUND_NBLK	= 6, /* We found node blk for attr */
> +	XFS_DAS_ALLOC_LEAF	= 7, /* We are allocating leaf blocks */
> +	XFS_DAS_FLIP_LFLAG	= 8, /* Flipped leaf INCOMPLETE attr flag */
> +	XFS_DAS_ALLOC_NODE	= 9, /* We are allocating node blocks */
> +	XFS_DAS_FLIP_NFLAG	= 10,/* Flipped node INCOMPLETE attr flag */

We've definitely reached the point where a state diagram would be
helpful.  Can you go from any of the RM_ states to the ones that you've
just added?

The new state machine code in last two patches would be a lot easier to
review if I could look down from above instead of up from the XFS_DAS
values and goto labels.  I /think/ it looks sane, but I'm only 20%
confident of that statement.

--D

>  };
>  
>  /*
> @@ -66,8 +74,13 @@ enum xfs_delattr_state {
>   * Context used for keeping track of delayed attribute operations
>   */
>  struct xfs_delattr_context {
> +	struct xfs_bmbt_irec	map;
> +	struct xfs_buf		*leaf_bp;
> +	xfs_fileoff_t		lfileoff;
>  	struct xfs_da_state	*da_state;
>  	struct xfs_da_state_blk *blk;
> +	xfs_dablk_t		lblkno;
> +	int			blkcnt;
>  	int			flags;
>  	enum xfs_delattr_state	dela_state;
>  };
> -- 
> 2.7.4
> 

Re: [PATCH v6 03/16] xfs: Add xfs_has_attr and subroutines

[Allison Collins]

On 1/21/20 3:48 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:22PM -0700, Allison Collins wrote:
>> From: Allison Henderson <allison.henderson@oracle.com>
>>
>> This patch adds a new functions to check for the existence of an
>> attribute.  Subroutines are also added to handle the cases of leaf
>> blocks, nodes or shortform.  Common code that appears in existing attr
>> add and remove functions have been factored out to help reduce the
>> appearance of duplicated code.  We will need these routines later for
>> delayed attributes since delayed operations cannot return error codes.
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c      | 171 ++++++++++++++++++++++++++++--------------
>>   fs/xfs/libxfs/xfs_attr.h      |   1 +
>>   fs/xfs/libxfs/xfs_attr_leaf.c | 111 +++++++++++++++++----------
>>   fs/xfs/libxfs/xfs_attr_leaf.h |   3 +
>>   4 files changed, 188 insertions(+), 98 deletions(-)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index 9acdb23..a2507b9 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -46,6 +46,7 @@ STATIC int xfs_attr_shortform_addname(xfs_da_args_t *args);
>>   STATIC int xfs_attr_leaf_get(xfs_da_args_t *args);
>>   STATIC int xfs_attr_leaf_addname(xfs_da_args_t *args);
>>   STATIC int xfs_attr_leaf_removename(xfs_da_args_t *args);
>> +STATIC int xfs_attr_leaf_hasname(struct xfs_da_args *args, struct xfs_buf **bp);
>>   
>>   /*
>>    * Internal routines when attribute list is more than one block.
>> @@ -53,6 +54,8 @@ STATIC int xfs_attr_leaf_removename(xfs_da_args_t *args);
>>   STATIC int xfs_attr_node_get(xfs_da_args_t *args);
>>   STATIC int xfs_attr_node_addname(xfs_da_args_t *args);
>>   STATIC int xfs_attr_node_removename(xfs_da_args_t *args);
>> +STATIC int xfs_attr_node_hasname(xfs_da_args_t *args,
>> +				 struct xfs_da_state **state);
>>   STATIC int xfs_attr_fillstate(xfs_da_state_t *state);
>>   STATIC int xfs_attr_refillstate(xfs_da_state_t *state);
>>   
>> @@ -310,6 +313,37 @@ xfs_attr_set_args(
>>   }
>>   
>>   /*
>> + * Return EEXIST if attr is found, or ENOATTR if not
>> + */
>> +int
>> +xfs_has_attr(
>> +	struct xfs_da_args      *args)
>> +{
>> +	struct xfs_inode	*dp = args->dp;
>> +	struct xfs_buf		*bp = NULL;
>> +	int			error;
>> +
>> +	if (!xfs_inode_hasattr(dp))
>> +		return -ENOATTR;
>> +
>> +	if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
>> +		ASSERT(dp->i_afp->if_flags & XFS_IFINLINE);
>> +		return xfs_attr_sf_findname(args, NULL, NULL);
>> +	}
>> +
>> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>> +		error = xfs_attr_leaf_hasname(args, &bp);
>> +
>> +		if (bp)
>> +			xfs_trans_brelse(args->trans, bp);
>> +
>> +		return error;
>> +	}
>> +
>> +	return xfs_attr_node_hasname(args, NULL);
>> +}
>> +
>> +/*
>>    * Remove the attribute specified in @args.
>>    */
>>   int
>> @@ -583,26 +617,20 @@ STATIC int
>>   xfs_attr_leaf_addname(
>>   	struct xfs_da_args	*args)
>>   {
>> -	struct xfs_inode	*dp;
>>   	struct xfs_buf		*bp;
>>   	int			retval, error, forkoff;
>> +	struct xfs_inode	*dp = args->dp;
>>   
>>   	trace_xfs_attr_leaf_addname(args);
>>   
>>   	/*
>> -	 * Read the (only) block in the attribute list in.
>> -	 */
>> -	dp = args->dp;
>> -	args->blkno = 0;
> 
> Don't we need this initialization?
I don't think so, it's set for use in the below xfs_attr3_leaf_read, but 
that line has been moved.  It's now over in xfs_attr_leaf_hasname. 
Initially I had the args->blkno = 0 assignment there too, but people 
thought it looked out of place and asked to remove it and just pass a 
hard 0 to the xfs_attr3_leaf_read call.


https://www.spinics.net/lists/linux-xfs/msg31863.html



> 
>> -	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
>> -	if (error)
>> -		return error;
>> -
>> -	/*
>>   	 * Look up the given attribute in the leaf block.  Figure out if
>>   	 * the given flags produce an error or call for an atomic rename.
>>   	 */
>> -	retval = xfs_attr3_leaf_lookup_int(bp, args);
>> +	retval = xfs_attr_leaf_hasname(args, &bp);
>> +	if (retval != -ENOATTR && retval != -EEXIST)
>> +		return retval;
>> +
>>   	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
>>   		xfs_trans_brelse(args->trans, bp);
>>   		return retval;
>> @@ -754,6 +782,27 @@ xfs_attr_leaf_addname(
>>   }
>>   
>>   /*
>> + * Return EEXIST if attr is found, or ENOATTR if not
>> + */
>> +STATIC int
>> +xfs_attr_leaf_hasname(
>> +	struct xfs_da_args      *args,
>> +	struct xfs_buf		**bp)
>> +{
>> +	int                     error = 0;
>> +
>> +	error = xfs_attr3_leaf_read(args->trans, args->dp, 0, bp);
>> +	if (error)
>> +		return error;
>> +
>> +	error = xfs_attr3_leaf_lookup_int(*bp, args);
>> +	if (error != -ENOATTR && error != -EEXIST)
>> +		xfs_trans_brelse(args->trans, *bp);
>> +
>> +	return error;
>> +}
>> +
>> +/*
>>    * Remove a name from the leaf attribute list structure
>>    *
>>    * This leaf block cannot have a "remote" value, we only call this routine
>> @@ -773,12 +822,11 @@ xfs_attr_leaf_removename(
>>   	 * Remove the attribute.
>>   	 */
>>   	dp = args->dp;
>> -	args->blkno = 0;
>> -	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
>> -	if (error)
>> +
>> +	error = xfs_attr_leaf_hasname(args, &bp);
>> +	if (error != -ENOATTR && error != -EEXIST)
>>   		return error;
>>   
>> -	error = xfs_attr3_leaf_lookup_int(bp, args);
>>   	if (error == -ENOATTR) {
>>   		xfs_trans_brelse(args->trans, bp);
>>   		return error;
>> @@ -817,12 +865,10 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
>>   
>>   	trace_xfs_attr_leaf_get(args);
>>   
>> -	args->blkno = 0;
>> -	error = xfs_attr3_leaf_read(args->trans, args->dp, args->blkno, &bp);
>> -	if (error)
>> +	error = xfs_attr_leaf_hasname(args, &bp);
>> +	if (error != -ENOATTR && error != -EEXIST)
>>   		return error;
>>   
>> -	error = xfs_attr3_leaf_lookup_int(bp, args);
>>   	if (error != -EEXIST)  {
>>   		xfs_trans_brelse(args->trans, bp);
>>   		return error;
>> @@ -832,6 +878,41 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
>>   	return error;
>>   }
>>   
>> +/*
>> + * Return EEXIST if attr is found, or ENOATTR if not
>> + * statep: If not null is set to point at the found state.  Caller will
>> + *         be responsible for freeing the state in this case.
> 
> Uh... a predicate returns an xfs_da_state object?
> 
> /me starts to wonder if these hasname functions ought to be some sort of
> lookup function that returns the appropriate context for inserting a new
> xattr entry if ENOATTR; a NULL context if EEXIST; or some other error if
> something actually went wrong...?

It's actually used in the remove path rather than the set path.  It 
looks like its used to reattach buffers after a commit has released 
them.  It's also one of the variables I need to stash in the delay 
context because it needs to persist across the transaction rolls.

> 
> Granted I have this funny feeling that most of the attr code works on
> the principle that we're supposed to pass ENOATTR/EEXIST from lower
> layers to higher layers, which makes me unsure if it makes more sense to
> stop abusing those magic error codes (and therefore having to switch()
> on every error return) or not.
Yes, initially I think they were coded on that principal, but it's one 
we have to avoid with this new approach for the reasons I explained in 
the last email concerning the .finish_item callback.

> 
> I dunno, I also had questions that I attached to the previous version of
> this patch, so... <shrug>
Yes, I hope I was able to answer those in the last email?  Let me know 
if you need more clarification.  Thanks!  :-)

Allison

> 
> --D
> 
>> + */
>> +STATIC int
>> +xfs_attr_node_hasname(
>> +	struct xfs_da_args	*args,
>> +	struct xfs_da_state	**statep)
>> +{
>> +	struct xfs_da_state	*state;
>> +	int			retval, error;
>> +
>> +	state = xfs_da_state_alloc();
>> +	state->args = args;
>> +	state->mp = args->dp->i_mount;
>> +
>> +	if (statep != NULL)
>> +		*statep = NULL;
>> +
>> +	/*
>> +	 * Search to see if name exists, and get back a pointer to it.
>> +	 */
>> +	error = xfs_da3_node_lookup_int(state, &retval);
>> +	if (error == 0) {
>> +		if (statep != NULL)
>> +			*statep = state;
>> +		return retval;
>> +	}
>> +
>> +	xfs_da_state_free(state);
>> +
>> +	return error;
>> +}
>> +
>>   /*========================================================================
>>    * External routines when attribute list size > geo->blksize
>>    *========================================================================*/
>> @@ -864,20 +945,17 @@ xfs_attr_node_addname(
>>   	dp = args->dp;
>>   	mp = dp->i_mount;
>>   restart:
>> -	state = xfs_da_state_alloc();
>> -	state->args = args;
>> -	state->mp = mp;
>> -
>>   	/*
>>   	 * Search to see if name already exists, and get back a pointer
>>   	 * to where it should go.
>>   	 */
>> -	error = xfs_da3_node_lookup_int(state, &retval);
>> -	if (error)
>> +	retval = xfs_attr_node_hasname(args, &state);
>> +	if (retval != -ENOATTR)
>>   		goto out;
>> +
>>   	blk = &state->path.blk[ state->path.active-1 ];
>>   	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
>> -	if ((args->name.type & ATTR_REPLACE) && (retval == -ENOATTR)) {
>> +	if (args->name.type & ATTR_REPLACE) {
>>   		goto out;
>>   	} else if (retval == -EEXIST) {
>>   		if (args->name.type & ATTR_CREATE)
>> @@ -1079,29 +1157,15 @@ xfs_attr_node_removename(
>>   {
>>   	struct xfs_da_state	*state;
>>   	struct xfs_da_state_blk	*blk;
>> -	struct xfs_inode	*dp;
>>   	struct xfs_buf		*bp;
>>   	int			retval, error, forkoff;
>> +	struct xfs_inode	*dp = args->dp;
>>   
>>   	trace_xfs_attr_node_removename(args);
>>   
>> -	/*
>> -	 * Tie a string around our finger to remind us where we are.
>> -	 */
>> -	dp = args->dp;
>> -	state = xfs_da_state_alloc();
>> -	state->args = args;
>> -	state->mp = dp->i_mount;
>> -
>> -	/*
>> -	 * Search to see if name exists, and get back a pointer to it.
>> -	 */
>> -	error = xfs_da3_node_lookup_int(state, &retval);
>> -	if (error || (retval != -EEXIST)) {
>> -		if (error == 0)
>> -			error = retval;
>> +	error = xfs_attr_node_hasname(args, &state);
>> +	if (error != -EEXIST)
>>   		goto out;
>> -	}
>>   
>>   	/*
>>   	 * If there is an out-of-line value, de-allocate the blocks.
>> @@ -1196,7 +1260,8 @@ xfs_attr_node_removename(
>>   	error = 0;
>>   
>>   out:
>> -	xfs_da_state_free(state);
>> +	if (state)
>> +		xfs_da_state_free(state);
>>   	return error;
>>   }
>>   
>> @@ -1316,31 +1381,23 @@ xfs_attr_node_get(xfs_da_args_t *args)
>>   {
>>   	xfs_da_state_t *state;
>>   	xfs_da_state_blk_t *blk;
>> -	int error, retval;
>> +	int error;
>>   	int i;
>>   
>>   	trace_xfs_attr_node_get(args);
>>   
>> -	state = xfs_da_state_alloc();
>> -	state->args = args;
>> -	state->mp = args->dp->i_mount;
>> -
>>   	/*
>>   	 * Search to see if name exists, and get back a pointer to it.
>>   	 */
>> -	error = xfs_da3_node_lookup_int(state, &retval);
>> -	if (error) {
>> -		retval = error;
>> -		goto out_release;
>> -	}
>> -	if (retval != -EEXIST)
>> +	error = xfs_attr_node_hasname(args, &state);
>> +	if (error != -EEXIST)
>>   		goto out_release;
>>   
>>   	/*
>>   	 * Get the value, local or "remote"
>>   	 */
>>   	blk = &state->path.blk[state->path.active - 1];
>> -	retval = xfs_attr3_leaf_getvalue(blk->bp, args);
>> +	error = xfs_attr3_leaf_getvalue(blk->bp, args);
>>   
>>   	/*
>>   	 * If not in a transaction, we have to release all the buffers.
>> @@ -1352,7 +1409,7 @@ xfs_attr_node_get(xfs_da_args_t *args)
>>   	}
>>   
>>   	xfs_da_state_free(state);
>> -	return retval;
>> +	return error;
>>   }
>>   
>>   /* Returns true if the attribute entry name is valid. */
>> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
>> index 35043db..ce7b039 100644
>> --- a/fs/xfs/libxfs/xfs_attr.h
>> +++ b/fs/xfs/libxfs/xfs_attr.h
>> @@ -153,6 +153,7 @@ int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
>>   		 unsigned char *value, int valuelen, int flags);
>>   int xfs_attr_set_args(struct xfs_da_args *args);
>>   int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
>> +int xfs_has_attr(struct xfs_da_args *args);
>>   int xfs_attr_remove_args(struct xfs_da_args *args);
>>   int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
>>   		  int flags, struct attrlist_cursor_kern *cursor);
>> diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
>> index cb5ef66..9d6b68c 100644
>> --- a/fs/xfs/libxfs/xfs_attr_leaf.c
>> +++ b/fs/xfs/libxfs/xfs_attr_leaf.c
>> @@ -654,18 +654,66 @@ xfs_attr_shortform_create(xfs_da_args_t *args)
>>   }
>>   
>>   /*
>> + * Return -EEXIST if attr is found, or -ENOATTR if not
>> + * args:  args containing attribute name and namelen
>> + * sfep:  If not null, pointer will be set to the last attr entry found on
>> +	  -EEXIST.  On -ENOATTR pointer is left at the last entry in the list
>> + * basep: If not null, pointer is set to the byte offset of the entry in the
>> + *	  list on -EEXIST.  On -ENOATTR, pointer is left at the byte offset of
>> + *	  the last entry in the list
>> + */
>> +int
>> +xfs_attr_sf_findname(
>> +	struct xfs_da_args	 *args,
>> +	struct xfs_attr_sf_entry **sfep,
>> +	unsigned int		 *basep)
>> +{
>> +	struct xfs_attr_shortform *sf;
>> +	struct xfs_attr_sf_entry *sfe;
>> +	unsigned int		base = sizeof(struct xfs_attr_sf_hdr);
>> +	int			size = 0;
>> +	int			end;
>> +	int			i;
>> +
>> +	sf = (struct xfs_attr_shortform *)args->dp->i_afp->if_u1.if_data;
>> +	sfe = &sf->list[0];
>> +	end = sf->hdr.count;
>> +	for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe),
>> +			base += size, i++) {
>> +		size = XFS_ATTR_SF_ENTSIZE(sfe);
>> +		if (sfe->namelen != args->name.len)
>> +			continue;
>> +		if (memcmp(sfe->nameval, args->name.name, args->name.len) != 0)
>> +			continue;
>> +		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
>> +			continue;
>> +		break;
>> +	}
>> +
>> +	if (sfep != NULL)
>> +		*sfep = sfe;
>> +
>> +	if (basep != NULL)
>> +		*basep = base;
>> +
>> +	if (i == end)
>> +		return -ENOATTR;
>> +	return -EEXIST;
>> +}
>> +
>> +/*
>>    * Add a name/value pair to the shortform attribute list.
>>    * Overflow from the inode has already been checked for.
>>    */
>>   void
>> -xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff)
>> +xfs_attr_shortform_add(struct xfs_da_args *args, int forkoff)
>>   {
>> -	xfs_attr_shortform_t *sf;
>> -	xfs_attr_sf_entry_t *sfe;
>> -	int i, offset, size;
>> -	xfs_mount_t *mp;
>> -	xfs_inode_t *dp;
>> -	struct xfs_ifork *ifp;
>> +	struct xfs_attr_shortform	*sf;
>> +	struct xfs_attr_sf_entry	*sfe;
>> +	int				offset, size, error;
>> +	struct xfs_mount		*mp;
>> +	struct xfs_inode		*dp;
>> +	struct xfs_ifork		*ifp;
>>   
>>   	trace_xfs_attr_sf_add(args);
>>   
>> @@ -676,18 +724,8 @@ xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff)
>>   	ifp = dp->i_afp;
>>   	ASSERT(ifp->if_flags & XFS_IFINLINE);
>>   	sf = (xfs_attr_shortform_t *)ifp->if_u1.if_data;
>> -	sfe = &sf->list[0];
>> -	for (i = 0; i < sf->hdr.count; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) {
>> -#ifdef DEBUG
>> -		if (sfe->namelen != args->name.len)
>> -			continue;
>> -		if (memcmp(args->name.name, sfe->nameval, args->name.len) != 0)
>> -			continue;
>> -		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
>> -			continue;
>> -		ASSERT(0);
>> -#endif
>> -	}
>> +	error = xfs_attr_sf_findname(args, &sfe, NULL);
>> +	ASSERT(error != -EEXIST);
>>   
>>   	offset = (char *)sfe - (char *)sf;
>>   	size = XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
>> @@ -730,35 +768,26 @@ xfs_attr_fork_remove(
>>    * Remove an attribute from the shortform attribute list structure.
>>    */
>>   int
>> -xfs_attr_shortform_remove(xfs_da_args_t *args)
>> +xfs_attr_shortform_remove(struct xfs_da_args *args)
>>   {
>> -	xfs_attr_shortform_t *sf;
>> -	xfs_attr_sf_entry_t *sfe;
>> -	int base, size=0, end, totsize, i;
>> -	xfs_mount_t *mp;
>> -	xfs_inode_t *dp;
>> +	struct xfs_attr_shortform	*sf;
>> +	struct xfs_attr_sf_entry	*sfe;
>> +	int				size = 0, end, totsize;
>> +	unsigned int			base;
>> +	struct xfs_mount		*mp;
>> +	struct xfs_inode		*dp;
>> +	int				error;
>>   
>>   	trace_xfs_attr_sf_remove(args);
>>   
>>   	dp = args->dp;
>>   	mp = dp->i_mount;
>> -	base = sizeof(xfs_attr_sf_hdr_t);
>>   	sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data;
>> -	sfe = &sf->list[0];
>> -	end = sf->hdr.count;
>> -	for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe),
>> -					base += size, i++) {
>> -		size = XFS_ATTR_SF_ENTSIZE(sfe);
>> -		if (sfe->namelen != args->name.len)
>> -			continue;
>> -		if (memcmp(sfe->nameval, args->name.name, args->name.len) != 0)
>> -			continue;
>> -		if (!xfs_attr_namesp_match(args->name.type, sfe->flags))
>> -			continue;
>> -		break;
>> -	}
>> -	if (i == end)
>> -		return -ENOATTR;
>> +
>> +	error = xfs_attr_sf_findname(args, &sfe, &base);
>> +	if (error != -EEXIST)
>> +		return error;
>> +	size = XFS_ATTR_SF_ENTSIZE(sfe);
>>   
>>   	/*
>>   	 * Fix up the attribute fork data, covering the hole
>> diff --git a/fs/xfs/libxfs/xfs_attr_leaf.h b/fs/xfs/libxfs/xfs_attr_leaf.h
>> index f4a188e..f6165ff 100644
>> --- a/fs/xfs/libxfs/xfs_attr_leaf.h
>> +++ b/fs/xfs/libxfs/xfs_attr_leaf.h
>> @@ -62,6 +62,9 @@ int	xfs_attr_shortform_getvalue(struct xfs_da_args *args);
>>   int	xfs_attr_shortform_to_leaf(struct xfs_da_args *args,
>>   			struct xfs_buf **leaf_bp);
>>   int	xfs_attr_shortform_remove(struct xfs_da_args *args);
>> +int	xfs_attr_sf_findname(struct xfs_da_args *args,
>> +			     struct xfs_attr_sf_entry **sfep,
>> +			     unsigned int *basep);
>>   int	xfs_attr_shortform_allfit(struct xfs_buf *bp, struct xfs_inode *dp);
>>   int	xfs_attr_shortform_bytesfit(struct xfs_inode *dp, int bytes);
>>   xfs_failaddr_t xfs_attr_shortform_verify(struct xfs_inode *ip);
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 06/16] xfs: Factor out xfs_attr_leaf_addname helper

[Allison Collins]

On 1/21/20 4:01 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:25PM -0700, Allison Collins wrote:
>> Factor out new helper function xfs_attr_leaf_try_add. Because new
>> delayed attribute routines cannot roll transactions, we carve off the
>> parts of xfs_attr_leaf_addname that we can use, and move the commit into
>> the calling function.
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> Reviewed-by: Brian Foster <bfoster@redhat.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c | 83 ++++++++++++++++++++++++++++--------------------
>>   1 file changed, 49 insertions(+), 34 deletions(-)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index b0ec25b..9ed7e94 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -305,10 +305,30 @@ xfs_attr_set_args(
>>   		}
>>   	}
>>   
>> -	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK))
>> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>>   		error = xfs_attr_leaf_addname(args);
>> -	else
>> -		error = xfs_attr_node_addname(args);
>> +		if (error != -ENOSPC)
>> +			return error;
>> +
>> +		/*
>> +		 * Commit that transaction so that the node_addname()
>> +		 * call can manage its own transactions.
>> +		 */
>> +		error = xfs_defer_finish(&args->trans);
>> +		if (error)
>> +			return error;
>> +
>> +		/*
>> +		 * Commit the current trans (including the inode) and
>> +		 * start a new one.
>> +		 */
>> +		error = xfs_trans_roll_inode(&args->trans, dp);
>> +		if (error)
>> +			return error;
>> +
>> +	}
>> +
>> +	error = xfs_attr_node_addname(args);
>>   	return error;
>>   }
>>   
>> @@ -607,21 +627,12 @@ xfs_attr_shortform_addname(xfs_da_args_t *args)
>>    * External routines when attribute list is one block
>>    *========================================================================*/
>>   
>> -/*
>> - * Add a name to the leaf attribute list structure
>> - *
>> - * This leaf block cannot have a "remote" value, we only call this routine
>> - * if bmap_one_block() says there is only one block (ie: no remote blks).
>> - */
> 
> These functions are complicated enough as they are now, please make sure
> they all have comments laying out the expected input metadata states and
> what output states result from them.
> 
> I /think/ this function takes an inode whose attr data is in leaf
> format, and tries to add a new entry.  If that succeeds then we exit
> with dirty inode and transaction.  If ENOSPC then we convert the attr
> data to node format and exit w/ dirty inode + transaction, presuming
> that the caller will try again with xfs_attr_node_addname?
That sounds about right.  How about the following commentary then:
/

/* 

  * Tries to add an attribute to an inode in leaf form 

  * 

  * This function is meant to execute as part of a delayed operation and 
leaves
  * the transaction handling to the caller.  On success the attribute is 
added
  * and the inode and transaction are left dirty.  If there is not 
enough space,
  * the attr data is converted to node format and -ENOSPC is returned. 

  * Caller is responsible for handling the dirty inode and transaction 
or adding
  * the attr in node format. 

  */

Sound good?


> 
>>   STATIC int
>> -xfs_attr_leaf_addname(
>> -	struct xfs_da_args	*args)
>> +xfs_attr_leaf_try_add(
>> +	struct xfs_da_args	*args,
>> +	struct xfs_buf		*bp)
>>   {
>> -	struct xfs_buf		*bp;
>> -	int			retval, error, forkoff;
>> -	struct xfs_inode	*dp = args->dp;
>> -
>> -	trace_xfs_attr_leaf_addname(args);
>> +	int			retval, error;
>>   
>>   	/*
>>   	 * Look up the given attribute in the leaf block.  Figure out if
>> @@ -667,31 +678,35 @@ xfs_attr_leaf_addname(
>>   	retval = xfs_attr3_leaf_add(bp, args);
>>   	if (retval == -ENOSPC) {
>>   		/*
>> -		 * Promote the attribute list to the Btree format, then
>> -		 * Commit that transaction so that the node_addname() call
>> -		 * can manage its own transactions.
>> +		 * Promote the attribute list to the Btree format.
>> +		 * Unless an error occurs, retain the -ENOSPC retval
>>   		 */
>>   		error = xfs_attr3_leaf_to_node(args);
>>   		if (error)
>>   			return error;
>> -		error = xfs_defer_finish(&args->trans);
>> -		if (error)
>> -			return error;
>> +	}
>> +	return retval;
>> +}
>>   
>> -		/*
>> -		 * Commit the current trans (including the inode) and start
>> -		 * a new one.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, dp);
>> -		if (error)
>> -			return error;
>>   
>> -		/*
>> -		 * Fob the whole rest of the problem off on the Btree code.
>> -		 */
>> -		error = xfs_attr_node_addname(args);
>> +/*
>> + * Add a name to the leaf attribute list structure
>> + *
>> + * This leaf block cannot have a "remote" value, we only call this routine
>> + * if bmap_one_block() says there is only one block (ie: no remote blks).
>> + */
>> +STATIC int
>> +xfs_attr_leaf_addname(struct xfs_da_args	*args)
> 
> STATIC int
> xfs_attr_leaf_addname(
> 	struct xfs_da_args	*args)
> {
> 
> Please fix the inconsistent style things as you touch them.
> 
> --D
Sure, will do.

Allison

> 
>> +{
>> +	int			error, forkoff;
>> +	struct xfs_buf		*bp = NULL;
>> +	struct xfs_inode	*dp = args->dp;
>> +
>> +	trace_xfs_attr_leaf_addname(args);
>> +
>> +	error = xfs_attr_leaf_try_add(args, bp);
>> +	if (error)
>>   		return error;
>> -	}
>>   
>>   	/*
>>   	 * Commit the transaction that added the attr name so that
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 07/16] xfs: Refactor xfs_attr_try_sf_addname

[Allison Collins]

On 1/21/20 4:07 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:26PM -0700, Allison Collins wrote:
>> To help pre-simplify xfs_attr_set_args, we need to hoist transacation
>> handling up, while modularizing the adjacent code down into helpers. In
>> this patch, hoist the commit in xfs_attr_try_sf_addname up into the
>> calling function, and also pull the attr list creation down.
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c | 30 +++++++++++++++---------------
>>   1 file changed, 15 insertions(+), 15 deletions(-)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index 9ed7e94..c15361a 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -227,8 +227,13 @@ xfs_attr_try_sf_addname(
>>   	struct xfs_da_args	*args)
>>   {
>>   
>> -	struct xfs_mount	*mp = dp->i_mount;
>> -	int			error, error2;
>> +	int			error;
>> +
>> +	/*
>> +	 * Build initial attribute list (if required).
>> +	 */
>> +	if (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS)
>> +		xfs_attr_shortform_create(args);
>>   
>>   	error = xfs_attr_shortform_addname(args);
>>   	if (error == -ENOSPC)
>> @@ -241,12 +246,10 @@ xfs_attr_try_sf_addname(
>>   	if (!error && (args->name.type & ATTR_KERNOTIME) == 0)
>>   		xfs_trans_ichgtime(args->trans, dp, XFS_ICHGTIME_CHG);
>>   
>> -	if (mp->m_flags & XFS_MOUNT_WSYNC)
>> +	if (dp->i_mount->m_flags & XFS_MOUNT_WSYNC)
>>   		xfs_trans_set_sync(args->trans);
>>   
>> -	error2 = xfs_trans_commit(args->trans);
>> -	args->trans = NULL;
>> -	return error ? error : error2;
>> +	return error;
>>   }
>>   
>>   /*
>> @@ -258,7 +261,7 @@ xfs_attr_set_args(
>>   {
>>   	struct xfs_inode	*dp = args->dp;
>>   	struct xfs_buf          *leaf_bp = NULL;
>> -	int			error;
>> +	int			error, error2 = 0;
>>   
>>   	/*
>>   	 * If the attribute list is non-existent or a shortform list,
>> @@ -269,17 +272,14 @@ xfs_attr_set_args(
>>   	     dp->i_d.di_anextents == 0)) {
>>   
>>   		/*
>> -		 * Build initial attribute list (if required).
>> -		 */
>> -		if (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS)
>> -			xfs_attr_shortform_create(args);
>> -
>> -		/*
>>   		 * Try to add the attr to the attribute list in the inode.
>>   		 */
>>   		error = xfs_attr_try_sf_addname(dp, args);
>> -		if (error != -ENOSPC)
>> -			return error;
>> +		if (error != -ENOSPC) {
>> +			error2 = xfs_trans_commit(args->trans);
> 
> /me wonders if there's really a point in committing the transaction for
> things like EEXIST and ENOMEM, but I guess this is a straight
> conversion and the current code really does this...
> 
> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
> 
> --D

Alrighty, thanks for the review!

Allison

> 
>> +			args->trans = NULL;
>> +			return error ? error : error2;
>> +		}
>>   
>>   		/*
>>   		 * It won't fit in the shortform, transform to a leaf block.
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 11/16] xfs: Check for -ENOATTR or -EEXIST

[Allison Collins]

On 1/21/20 4:15 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:30PM -0700, Allison Collins wrote:
>> Delayed operations cannot return error codes.  So we must check for
>> these conditions first before starting set or remove operations
> 
> Answering my own question from earlier -- I see here you actually /are/
> checking the attr existence w.r.t. ATTR_{CREATE,REPLACE} right after we
> allocate a transaction and ILOCK the inode, so
> 
> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Alrighty, thank you!

> 
> Though I am wondering if you could discard the predicates from the
> second patch in favor of doing a normal lookup of the attr with a zero
> valuelen to determine if there's already an attribute?
I think I likely answered this in the response to that patch.  Because 
it's used as part of the remove procedures, we still need it.  We could 
make a simpler version just for this application I suppose, but it seems 
like it'd just be extra code since we still need the former.

Thank you for the reviews!
Allison

> 
> --D
> 
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c | 12 ++++++++++++
>>   1 file changed, 12 insertions(+)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index a2673fe..e9d22c1 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -457,6 +457,14 @@ xfs_attr_set(
>>   		goto out_trans_cancel;
>>   
>>   	xfs_trans_ijoin(args.trans, dp, 0);
>> +
>> +	error = xfs_has_attr(&args);
>> +	if (error == -EEXIST && (name->type & ATTR_CREATE))
>> +		goto out_trans_cancel;
>> +
>> +	if (error == -ENOATTR && (name->type & ATTR_REPLACE))
>> +		goto out_trans_cancel;
>> +
>>   	error = xfs_attr_set_args(&args);
>>   	if (error)
>>   		goto out_trans_cancel;
>> @@ -545,6 +553,10 @@ xfs_attr_remove(
>>   	 */
>>   	xfs_trans_ijoin(args.trans, dp, 0);
>>   
>> +	error = xfs_has_attr(&args);
>> +	if (error != -EEXIST)
>> +		goto out;
>> +
>>   	error = xfs_attr_remove_args(&args);
>>   	if (error)
>>   		goto out;
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 12/16] xfs: Add helper function xfs_attr_init_unmapstate

[Allison Collins]

On 1/21/20 4:21 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:31PM -0700, Allison Collins wrote:
>> This patch helps to pre-simplify xfs_attr_node_removename by modularizing
>> the code around the transactions into helper functions.  This will make
>> the function easier to follow when we introduce delayed attributes.
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c | 45 +++++++++++++++++++++++++++++++--------------
>>   1 file changed, 31 insertions(+), 14 deletions(-)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index e9d22c1..453ea59 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -1202,6 +1202,36 @@ xfs_attr_node_addname(
>>   }
>>   
>>   /*
>> + * Set up the state for unmap and set the incomplete flag
> 
> "Mark an attribute entry INCOMPLETE and save pointers to the relevant
> buffers for later deletion of the entry." ?
I am fine with that if folks prefer.

> 
>> + */
>> +STATIC int
>> +xfs_attr_init_unmapstate(
> 
> I'm hung up on this name -- it marks the entry incomplete and saves some
> breadcrumbs for erasing data later, which I guess is "unmap state"?
> What do you think of:
> 
> xfs_attr_leaf_mark_incomplete()

I actually struggled a bit with what to call it.  I called it an init 
because in the immediately proceeding code we fall into a loop of 
bunmapi and refilling the state, rolling the transaction in between. 
But we only make this initial call once before we start doing that. 
Later when we get into the state machine this turns into multiple 
iterations of jumping in and out of this region of code.  Even though 
the while() syntax is gone, it still ends up functioning very much like 
a loop.

This helper function isnt so much necessary as much as it is an effort 
to cut down on the size of xfs_attr_node_removename which was getting 
quite lengthy.

In anycase, xfs_attr_leaf_mark_incomplete is fine with me :-)

> 
>> +	struct xfs_da_args	*args,
>> +	struct xfs_da_state	*state)
>> +{
>> +	int error;
>> +
>> +	/*
>> +	 * Fill in disk block numbers in the state structure
>> +	 * so that we can get the buffers back after we commit
>> +	 * several transactions in the following calls.
>> +	 */
>> +	error = xfs_attr_fillstate(state);
>> +	if (error)
>> +		return error;
>> +
>> +	/*
>> +	 * Mark the attribute as INCOMPLETE
> 
> Has this stopped doing the bunmapi call, or was the comment inaccurate?
> I'm guessing the second if it's necessary to save state to recover
> buffer pointers later...?
Right, I clipped off the comment because we dont do the bunmapi in this 
function scope.  That happens later in the loop behavior described above.

Hope that helps!

Allison

> 
> --D
> 
>> +	 */
>> +	error = xfs_attr3_leaf_setflag(args);
>> +	if (error)
>> +		return error;
>> +
>> +	return 0;
>> +}
>> +
>> +
>> +/*
>>    * Remove a name from a B-tree attribute list.
>>    *
>>    * This will involve walking down the Btree, and may involve joining
>> @@ -1233,20 +1263,7 @@ xfs_attr_node_removename(
>>   	ASSERT(blk->bp != NULL);
>>   	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
>>   	if (args->rmtblkno > 0) {
>> -		/*
>> -		 * Fill in disk block numbers in the state structure
>> -		 * so that we can get the buffers back after we commit
>> -		 * several transactions in the following calls.
>> -		 */
>> -		error = xfs_attr_fillstate(state);
>> -		if (error)
>> -			goto out;
>> -
>> -		/*
>> -		 * Mark the attribute as INCOMPLETE, then bunmapi() the
>> -		 * remote value.
>> -		 */
>> -		error = xfs_attr3_leaf_setflag(args);
>> +		error = xfs_attr_init_unmapstate(args, state);
>>   		if (error)
>>   			goto out;
>>   
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 13/16] xfs: Add helper function xfs_attr_rmtval_unmap

[Allison Collins]

On 1/21/20 4:24 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:32PM -0700, Allison Collins wrote:
>> This function is similar to xfs_attr_rmtval_remove, but adapted to
>> return. EAGAIN for new transactions. We will use this later when we
> 
> "..to return -EAGAIN"?
yees, I noticed that after I sent it... will fix :-)
> 
>> introduce delayed attributes
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr_remote.c | 27 +++++++++++++++++++++++++++
>>   fs/xfs/libxfs/xfs_attr_remote.h |  1 +
>>   2 files changed, 28 insertions(+)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
>> index 9b4fa2a..f2ee0b8 100644
>> --- a/fs/xfs/libxfs/xfs_attr_remote.c
>> +++ b/fs/xfs/libxfs/xfs_attr_remote.c
>> @@ -684,3 +684,30 @@ xfs_attr_rmtval_remove(
>>   	}
>>   	return 0;
>>   }
>> +
>> +/*
>> + * Unmap value blocks for this attr.  This is similar to
>> + * xfs_attr_rmtval_remove, but adapted to to return EAGAIN for new transactions
note to self: fix too many to's

> 
> Do you have to scan for and invalidate any incore buffers for the remote
> value?  Or will that be performed by another step?  Hard to tell because
> this function doesn't have any callers.
Yes, that happens in xfs_attr_rmtval_invalidate.  Right after we mark 
incomplete, and before we fall into the bunmapi loop I mentioned in the 
last patch.  Basically this function becomes the inside of that loop.


> 
>> + */
>> +int
>> +xfs_attr_rmtval_unmap(
>> +	struct xfs_da_args	*args)
>> +{
>> +	int	error, done;
>> +
>> +	/*
>> +	 * Unmap value blocks for this attr.  This is similar to
>> +	 * xfs_attr_rmtval_remove, but open coded here to return EAGAIN
>> +	 * for new transactions
>> +	 */
>> +	error = xfs_bunmapi(args->trans, args->dp,
>> +		    args->rmtblkno, args->rmtblkcnt,
>> +		    XFS_BMAPI_ATTRFORK, 1, &done);
>> +	if (error)
>> +		return error;
>> +
>> +	if (!done)
>> +		return -EAGAIN;
> 
> return done ? 0 : -EAGAIN; ?
I believe Christoph asked for the return 0;  I am not particular about 
it as long a people are in agreement.  If people feel strongly one way 
or another please chime in.

Allison
> 
> --D
> 


>> +
>> +	return 0;
>> +}
>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.h b/fs/xfs/libxfs/xfs_attr_remote.h
>> index 85f2593..7ab3770 100644
>> --- a/fs/xfs/libxfs/xfs_attr_remote.h
>> +++ b/fs/xfs/libxfs/xfs_attr_remote.h
>> @@ -12,4 +12,5 @@ int xfs_attr_rmtval_get(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_set(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_remove(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
>> +int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
>>   #endif /* __XFS_ATTR_REMOTE_H__ */
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 14/16] xfs: Simplify xfs_attr_set_args

[Allison Collins]

On 1/21/20 4:31 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:33PM -0700, Allison Collins wrote:
>> Delayed attribute mechanics make frequent use of goto statements.  We
>> can use this to further simplify xfs_attr_set_args.  In this patch we
>> introduce one of the gotos now to help pre-simplify the routine.  This
>> will help make proceeding patches in this area easier to read.
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c | 71 +++++++++++++++++++++++++-----------------------
>>   1 file changed, 37 insertions(+), 34 deletions(-)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index 453ea59..90e0b2d 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -264,47 +264,50 @@ xfs_attr_set_args(
>>   	int			error, error2 = 0;
>>   
>>   	/*
>> -	 * If the attribute list is non-existent or a shortform list,
>> -	 * upgrade it to a single-leaf-block attribute list.
>> +	 * If the attribute list is non-existent or a shortform list, proceed to
>> +	 * upgrade it to a single-leaf-block attribute list.  Otherwise jump
>> +	 * straight to leaf handling
> 
> We don't jump to "upgrade it to a single leaf block attr list" quite
> that soon.
> 
> "If the attribute list is already in leaf format, jump straight to leaf
> handling.  Otherwise, try to add the attribute to the shortform list; if
> there's no room then convert the list to leaf format and try again." ?
Sure, I think that sounds good.

> 
>>   	 */
>> -	if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL ||
>> +	if (!(dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL ||
>>   	    (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
>> -	     dp->i_d.di_anextents == 0)) {
>> +	     dp->i_d.di_anextents == 0)))
> 
> Also... maybe this should just be a separate predicate?
> 
> static inline bool
> xfs_attr_is_leaf_format(
> 	struct xfs_inode	*dp)
> {
> 	return dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
> 		dp->i_d.di_anextents > 0;
> }
> 
> 
> if (xfs_attr_is_leaf_format(dp))
> 	goto add_leaf;

Ok, I think that looks better too.  Will add.

> 
>> +		goto sf_to_leaf;
>>   
>> -		/*
>> -		 * Try to add the attr to the attribute list in the inode.
>> -		 */
>> -		error = xfs_attr_try_sf_addname(dp, args);
>> -		if (error != -ENOSPC) {
>> -			error2 = xfs_trans_commit(args->trans);
>> -			args->trans = NULL;
>> -			return error ? error : error2;
>> -		}
>> +	/*
>> +	 * Try to add the attr to the attribute list in the inode.
>> +	 */
>> +	error = xfs_attr_try_sf_addname(dp, args);
>> +	if (error != -ENOSPC) {
>> +		error2 = xfs_trans_commit(args->trans);
>> +		args->trans = NULL;
>> +		return error ? error : error2;
>> +	}
>>   
>> -		/*
>> -		 * It won't fit in the shortform, transform to a leaf block.
>> -		 * GROT: another possible req'mt for a double-split btree op.
>> -		 */
>> -		error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
>> -		if (error)
>> -			return error;
>> +	/*
>> +	 * It won't fit in the shortform, transform to a leaf block.
>> +	 * GROT: another possible req'mt for a double-split btree op.
>> +	 */
>> +	error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
>> +	if (error)
>> +		return error;
>>   
>> -		/*
>> -		 * Prevent the leaf buffer from being unlocked so that a
>> -		 * concurrent AIL push cannot grab the half-baked leaf
>> -		 * buffer and run into problems with the write verifier.
>> -		 * Once we're done rolling the transaction we can release
>> -		 * the hold and add the attr to the leaf.
>> -		 */
>> -		xfs_trans_bhold(args->trans, leaf_bp);
>> -		error = xfs_defer_finish(&args->trans);
>> -		xfs_trans_bhold_release(args->trans, leaf_bp);
>> -		if (error) {
>> -			xfs_trans_brelse(args->trans, leaf_bp);
>> -			return error;
>> -		}
>> +	/*
>> +	 * Prevent the leaf buffer from being unlocked so that a
>> +	 * concurrent AIL push cannot grab the half-baked leaf
>> +	 * buffer and run into problems with the write verifier.
>> +	 * Once we're done rolling the transaction we can release
>> +	 * the hold and add the attr to the leaf.
>> +	 */
>> +	xfs_trans_bhold(args->trans, leaf_bp);
>> +	error = xfs_defer_finish(&args->trans);
>> +	xfs_trans_bhold_release(args->trans, leaf_bp);
>> +	if (error) {
>> +		xfs_trans_brelse(args->trans, leaf_bp);
>> +		return error;
>>   	}
>>   
>> +sf_to_leaf:
> 
> The attr list already has to be in list format by the time it gets here,
> which means that the label name is misleading.  How about "add_leaf" ?
Sure, I'll update the state machine names to match too then.

Thanks for the review!
Allison

> 
> --D
> 
>>   	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>>   		error = xfs_attr_leaf_addname(args);
>>   		if (error != -ENOSPC)
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 15/16] xfs: Add delay ready attr remove routines

[Allison Collins]

On 1/21/20 5:02 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:34PM -0700, Allison Collins wrote:
>> This patch modifies the attr remove routines to be delay ready. This
>> means they no longer roll or commit transactions, but instead return
>> -EAGAIN to have the calling routine roll and refresh the transaction.
>> In this series, xfs_attr_remove_args has become xfs_attr_remove_iter,
>> which uses a sort of state machine like switch to keep track of where it
>> was when EAGAIN was returned. xfs_attr_node_removename has also been
>> modified to use the switch, and a  new version of xfs_attr_remove_args
>> consists of a simple loop to refresh the transaction until the operation
>> is completed.  A helper function xfs_attr_node_shrink has also been
>> added to help simplify xfs_attr_node_removename and reduce length.
>>
>> This patch also adds a new struct xfs_delattr_context, which we will use
>> to keep track of the current state of an attribute operation. The new
>> xfs_delattr_state enum is used to track various operations that are in
>> progress so that we know not to repeat them, and resume where we left
>> off before EAGAIN was returned to cycle out the transaction. Other
>> members take the place of local variables that need to retain their
>> values across multiple function recalls.
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c     | 181 +++++++++++++++++++++++++++++++++----------
>>   fs/xfs/libxfs/xfs_attr.h     |   1 +
>>   fs/xfs/libxfs/xfs_da_btree.h |  24 ++++++
>>   fs/xfs/scrub/common.c        |   2 +
>>   fs/xfs/xfs_acl.c             |   2 +
>>   fs/xfs/xfs_attr_list.c       |   1 +
>>   fs/xfs/xfs_ioctl.c           |   2 +
>>   fs/xfs/xfs_ioctl32.c         |   2 +
>>   fs/xfs/xfs_iops.c            |   2 +
>>   fs/xfs/xfs_xattr.c           |   1 +
>>   10 files changed, 176 insertions(+), 42 deletions(-)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index 90e0b2d..92929ad 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -371,11 +371,60 @@ xfs_has_attr(
>>    */
>>   int
>>   xfs_attr_remove_args(
>> +	struct xfs_da_args	*args)
>> +{
>> +	int			error = 0;
>> +	int			err2 = 0;
>> +
>> +	do {
>> +		error = xfs_attr_remove_iter(args);
>> +		if (error && error != -EAGAIN)
>> +			goto out;
>> +
>> +		if (args->dac.flags & XFS_DAC_FINISH_TRANS) {
>> +			args->dac.flags &= ~XFS_DAC_FINISH_TRANS;
> 
> Won't rolling the transaction take care of calling defer_finish?
> 
> (Granted, I think you started this patch set long before we actually
> introduced that behavior...)

I actually plumbed in this extra flag field because folks has asked me 
to preserve the finish calls where they appeared in the last review. 
Reason being that this changes behavior in that the transaction roll 
doesn't complete pending deferred operations.

https://www.spinics.net/lists/linux-xfs/msg35096.html

> 
>> +			err2 = xfs_defer_finish(&args->trans);
>> +			if (err2) {
>> +				error = err2;
>> +				goto out;
>> +			}
>> +		}
>> +
>> +		err2 = xfs_trans_roll_inode(&args->trans, args->dp);
>> +		if (err2) {
>> +			error = err2;
>> +			goto out;
>> +		}
>> +
>> +	} while (error == -EAGAIN);
> 
> /me suspects this could be compressed down to:
> 
> do {
> 	error = xfs_attr_remove_iter(args);
> 	if (error == -EAGAIN) {
> 		error = 0;
> 		break;
?? if error is EAGAIN, we want to roll and continue, right?
And if it's 0 we roll and break.

> 	}
> 	if (error)
> 		break;
> 
> 	error = xfs_trans_roll_inode(...);
> } while (error == 0);
Logic bugs aside, I think you meant something pretty close to the v5 
version.

> 
> return error;
> 
>> +out:
>> +	return error;
>> +}
>> +
>> +/*
>> + * Remove the attribute specified in @args.
>> + * This routine is meant to function as a delayed operation, and may return
>> + * -EGAIN when the transaction needs to be rolled.  Calling functions will need
>> + * to handle this, and recall the function until a successful error code is
>> + * returned.
> 
> "This function may return -EAGAIN to signal that the transaction needs
> to be rolled.  Callers should continue calling this function until they
> receive a return value other than -EAGAIN." ?
Sure that's fine if folks think that is more clear.  At this point I 
feel like I have enough of it in my head, it's more about whether it 
makes sense to other people.  I agree the paraphrasing it equivalent.

> 
>> + */
>> +int
>> +xfs_attr_remove_iter(
>>   	struct xfs_da_args      *args)
>>   {
>>   	struct xfs_inode	*dp = args->dp;
>>   	int			error;
>>   
>> +	/* State machine switch */
>> +	switch (args->dac.dela_state) {
>> +	case XFS_DAS_RM_SHRINK:
>> +	case XFS_DAS_RM_NODE_BLKS:
>> +		goto node;
>> +	default:
>> +		break;
>> +	}
>> +
>>   	if (!xfs_inode_hasattr(dp)) {
>>   		error = -ENOATTR;
>>   	} else if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
>> @@ -384,6 +433,7 @@ xfs_attr_remove_args(
>>   	} else if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>>   		error = xfs_attr_leaf_removename(args);
>>   	} else {
>> +node:
>>   		error = xfs_attr_node_removename(args);
>>   	}
>>   
>> @@ -887,9 +937,8 @@ xfs_attr_leaf_removename(
>>   		/* bp is gone due to xfs_da_shrink_inode */
>>   		if (error)
>>   			return error;
>> -		error = xfs_defer_finish(&args->trans);
>> -		if (error)
>> -			return error;
>> +
>> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>   	}
>>   	return 0;
>>   }
>> @@ -1233,6 +1282,42 @@ xfs_attr_init_unmapstate(
>>   	return 0;
>>   }
>>   
>> +/*
>> + * Shrink an attribute from leaf to shortform
>> + */
>> +STATIC int
>> +xfs_attr_node_shrink(
>> +	struct xfs_da_args	*args,
>> +	struct xfs_da_state     *state)
>> +{
>> +	struct xfs_inode	*dp = args->dp;
>> +	int			error, forkoff;
>> +	struct xfs_buf		*bp;
>> +
>> +	/*
>> +	 * Have to get rid of the copy of this dabuf in the state.
>> +	 */
>> +	ASSERT(state->path.active == 1);
>> +	ASSERT(state->path.blk[0].bp);
>> +	state->path.blk[0].bp = NULL;
>> +
>> +	error = xfs_attr3_leaf_read(args->trans, args->dp, 0, &bp);
>> +	if (error)
>> +		return error;
>> +
>> +	forkoff = xfs_attr_shortform_allfit(bp, dp);
>> +	if (forkoff) {
>> +		error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
>> +		/* bp is gone due to xfs_da_shrink_inode */
>> +		if (error)
>> +			return error;
>> +
>> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +	} else
>> +		xfs_trans_brelse(args->trans, bp);
>> +
>> +	return 0;
>> +}
>>   
>>   /*
>>    * Remove a name from a B-tree attribute list.
>> @@ -1240,6 +1325,11 @@ xfs_attr_init_unmapstate(
>>    * This will involve walking down the Btree, and may involve joining
>>    * leaf nodes and even joining intermediate nodes up to and including
>>    * the root node (a special case of an intermediate node).
>> + *
>> + * This routine is meant to function as either an inline or delayed operation,
>> + * and may return -EAGAIN when the transaction needs to be rolled.  Calling
>> + * functions will need to handle this, and recall the function until a
>> + * successful error code is returned.
>>    */
>>   STATIC int
>>   xfs_attr_node_removename(
>> @@ -1247,15 +1337,28 @@ xfs_attr_node_removename(
>>   {
>>   	struct xfs_da_state	*state;
>>   	struct xfs_da_state_blk	*blk;
>> -	struct xfs_buf		*bp;
>> -	int			retval, error, forkoff;
>> +	int			retval, error;
>>   	struct xfs_inode	*dp = args->dp;
>>   
>>   	trace_xfs_attr_node_removename(args);
>> +	state = args->dac.da_state;
>> +	blk = args->dac.blk;
>> +
>> +	/* State machine switch */
>> +	switch (args->dac.dela_state) {
>> +	case XFS_DAS_RM_NODE_BLKS:
>> +		goto rm_node_blks;
>> +	case XFS_DAS_RM_SHRINK:
>> +		goto rm_shrink;
>> +	default:
>> +		break;
>> +	}
>>   
>>   	error = xfs_attr_node_hasname(args, &state);
>>   	if (error != -EEXIST)
>>   		goto out;
>> +	else
>> +		error = 0;
>>   
>>   	/*
>>   	 * If there is an out-of-line value, de-allocate the blocks.
>> @@ -1265,18 +1368,36 @@ xfs_attr_node_removename(
>>   	blk = &state->path.blk[ state->path.active-1 ];
>>   	ASSERT(blk->bp != NULL);
>>   	ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
>> +
>> +	/*
>> +	 * Store blk and state in the context incase we need to cycle out the
>> +	 * transaction
>> +	 */
>> +	args->dac.blk = blk;
>> +	args->dac.da_state = state;
>> +
>>   	if (args->rmtblkno > 0) {
>>   		error = xfs_attr_init_unmapstate(args, state);
>>   		if (error)
>>   			goto out;
>>   
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>> +		error = xfs_attr_rmtval_invalidate(args);
>>   		if (error)
>>   			goto out;
>> +	}
>>   
>> -		error = xfs_attr_rmtval_remove(args);
>> -		if (error)
>> -			goto out;
>> +rm_node_blks:
>> +
>> +	args->dac.dela_state = XFS_DAS_RM_NODE_BLKS;
> 
> That's odd -- we jumped here if dela_state == RM_NODE_BLKS, so why do we
> need to set dela_state?  Does that put us at risk of jumping back here
> by accident?
It would if there was a subroutine using the state machine, but we dont 
have that in this case, so it actually doesn't have an effect.  It was 
asked for in one of the reviews, I assumed because people found it 
appropriate to set it at that time, sort of like a witness mark.  It 
doesnt hurt anything to do so, but you are right in that it is a 
duplicate.

> 
> Oh, maybe this is supposed to make sure we keep coming back for more
> bunmapi if the the remote value blocks were backed by multiple extents.
> 
> Perhaps this would be better named DAS_RMTVAL_REMOVE?
xfs_attr_rmtval_unmap is the association in this case. 
DAS_RMTVAL_REMOVE is appropriate if folks prefer.

> 
>> +
>> +	if (args->rmtblkno > 0) {
>> +		error = xfs_attr_rmtval_unmap(args);
>> +
>> +		if (error) {
>> +			if (error == -EAGAIN)
>> +				args->dac.dela_state = XFS_DAS_RM_NODE_BLKS;
> 
> (Didn't we already set this state?)
It is.  If you prefer to only keep one, I would stick with this one 
because this is the specific case that really needs it.  Also I think 
having them there helps to mark which of the returns are sort of the 
"jump out" points

> 
>> +			return error;
>> +		}
>>   
>>   		/*
>>   		 * Refill the state structure with buffers, the prior calls
>> @@ -1302,44 +1423,20 @@ xfs_attr_node_removename(
>>   		error = xfs_da3_join(state);
>>   		if (error)
>>   			goto out;
>> -		error = xfs_defer_finish(&args->trans);
>> -		if (error)
>> -			goto out;
>> -		/*
>> -		 * Commit the Btree join operation and start a new trans.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, dp);
>> -		if (error)
>> -			goto out;
>> +
>> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +		args->dac.dela_state = XFS_DAS_RM_SHRINK;
>> +		return -EAGAIN;
>>   	}
>>   
>> +rm_shrink:
>> +	args->dac.dela_state = XFS_DAS_RM_SHRINK;
>> +
>>   	/*
>>   	 * If the result is small enough, push it all into the inode.
>>   	 */
>> -	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>> -		/*
>> -		 * Have to get rid of the copy of this dabuf in the state.
>> -		 */
>> -		ASSERT(state->path.active == 1);
>> -		ASSERT(state->path.blk[0].bp);
>> -		state->path.blk[0].bp = NULL;
>> -
>> -		error = xfs_attr3_leaf_read(args->trans, args->dp, 0, &bp);
>> -		if (error)
>> -			goto out;
>> -
>> -		if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
>> -			error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
>> -			/* bp is gone due to xfs_da_shrink_inode */
>> -			if (error)
>> -				goto out;
>> -			error = xfs_defer_finish(&args->trans);
>> -			if (error)
>> -				goto out;
>> -		} else
>> -			xfs_trans_brelse(args->trans, bp);
>> -	}
>> -	error = 0;
>> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK))
>> +		error = xfs_attr_node_shrink(args, state);
>>   
>>   out:
>>   	if (state)
>> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
>> index ce7b039..ea873a5 100644
>> --- a/fs/xfs/libxfs/xfs_attr.h
>> +++ b/fs/xfs/libxfs/xfs_attr.h
>> @@ -155,6 +155,7 @@ int xfs_attr_set_args(struct xfs_da_args *args);
>>   int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
>>   int xfs_has_attr(struct xfs_da_args *args);
>>   int xfs_attr_remove_args(struct xfs_da_args *args);
>> +int xfs_attr_remove_iter(struct xfs_da_args *args);
>>   int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
>>   		  int flags, struct attrlist_cursor_kern *cursor);
>>   bool xfs_attr_namecheck(const void *name, size_t length);
>> diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
>> index 14f1be3..7fc87da 100644
>> --- a/fs/xfs/libxfs/xfs_da_btree.h
>> +++ b/fs/xfs/libxfs/xfs_da_btree.h
>> @@ -50,9 +50,33 @@ enum xfs_dacmp {
>>   };
>>   
>>   /*
>> + * Enum values for xfs_delattr_context.da_state
>> + */
> 
> I can tell this is an enum. ;)
> 
> Could you have this comment explain a little more about how states are
> managed?  I /think/ these state values mark where we were in the attr
> code when something decided it was time to roll a transaction, right?
> 
> In other words, dela_state is a save point where we jump back to after
> hopping out to roll a transaction?
> 
> (The parts of the code we jump somewhere based on dela_state and
> immediately set dela_state confused me, so I don't know if I've
> understood how the state machine works correctly.)
It sounds like you're pretty much understanding it.  If that extra state 
set was confusing, then perhaps it should go.  Anyway, how about the 
following commentary:

/*
  * Enum values for xfs_delattr_context.da_state
  *
  * These values are used by delayed attribute operations to keep track
  * of where they were before they returned -EAGAIN.  A return code of
  * -EAGAIN signals the calling function to roll the transaction, and
  * then recall the subroutine to finish the operation.  The enum
  * is then used by the subroutine to jump back to where it was and
  * resume executing where it left off.
  */

> 
>> +enum xfs_delattr_state {
>> +	XFS_DAS_RM_SHRINK	= 1, /* We are shrinking the tree */
> 
> This means that dela_state == 0 means "do not use delayed attributes",
> right?  We might want to make the meaning of zero more explicit.
No, at this point a value of 0 does not mean anything.  It will later 
when we get into delayed attrs.  Delayed attrs are slightly different in 
that args (and the state machine) need to get set up after the "ping 
pong" has already started.  So we initialize on a state of 0, and have 
one more extra state (XFS_DAS_INIT) just to say that we did it, so dont 
initialize it again.  (See xfs_attr_finish_item from the git branch I 
sent earlier)

To further comment your interpretation, no, the routines do not execute 
any differently when the state is zero, or any other value they do not 
reconize.  They simply assume to "start from the begining" (ie: they 
simply fall through the switch statement) and over write the state as 
needed.

In fact, there are a few spots in the next patch when a child function 
gets a state reminicent of it's parent.  This condidtion is not an 
error.  It just means that the parent function was using it up until 
this point when it choose  to surrender the statemachine to the child 
function.  The child function has no notion of what its parents states 
even mean and pay no mind.  It is however the parents responsibility to 
reconize states of its child, and jump back to the child function to 
allow it to finish its tasks.

Hope that helps some :-)

> 
>> +	XFS_DAS_RM_NODE_BLKS	= 2, /* We are removing node blocks */
> 
> Also, the C compiler will auto-increment this for you if you don't
> assign an explicit value.
Ok, do you prefer them not to be explicitly assigned?

> 
>> +};
>> +
>> +/*
>> + * Defines for xfs_delattr_context.flags
>> + */
>> +#define	XFS_DAC_FINISH_TRANS	0x1 /* indicates to finish the transaction */
>> +
>> +/*
>> + * Context used for keeping track of delayed attribute operations
>> + */
>> +struct xfs_delattr_context {
>> +	struct xfs_da_state	*da_state;
>> +	struct xfs_da_state_blk *blk;
>> +	int			flags;
> 
> unsigned int, assuming FINISH_TRANS doesn't just disappear.
Sure, will fix

> 
>> +	enum xfs_delattr_state	dela_state;
>> +};
>> +
>> +/*
>>    * Structure to ease passing around component names.
>>    */
>>   typedef struct xfs_da_args {
>> +	struct xfs_delattr_context dac; /* context used for delay attr ops */
>>   	struct xfs_da_geometry *geo;	/* da block geometry */
>>   	struct xfs_name	name;		/* name, length and argument  flags*/
>>   	uint8_t		filetype;	/* filetype of inode for directories */
>> diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
>> index 1887605..9a649d1 100644
>> --- a/fs/xfs/scrub/common.c
>> +++ b/fs/xfs/scrub/common.c
>> @@ -24,6 +24,8 @@
>>   #include "xfs_rmap_btree.h"
>>   #include "xfs_log.h"
>>   #include "xfs_trans_priv.h"
>> +#include "xfs_da_format.h"
>> +#include "xfs_da_btree.h"
> 
> Are these necessary if all the da_args stuff moves to xfs_types.h?
That may remedy some of this.  I'll see if I can move it over.  Thanks 
for the review.

Allison

> 
> --D
> 
>>   #include "xfs_attr.h"
>>   #include "xfs_reflink.h"
>>   #include "scrub/scrub.h"
>> diff --git a/fs/xfs/xfs_acl.c b/fs/xfs/xfs_acl.c
>> index 42ac847..d65e6d8 100644
>> --- a/fs/xfs/xfs_acl.c
>> +++ b/fs/xfs/xfs_acl.c
>> @@ -10,6 +10,8 @@
>>   #include "xfs_trans_resv.h"
>>   #include "xfs_mount.h"
>>   #include "xfs_inode.h"
>> +#include "xfs_da_format.h"
>> +#include "xfs_da_btree.h"
>>   #include "xfs_attr.h"
>>   #include "xfs_trace.h"
>>   #include "xfs_error.h"
>> diff --git a/fs/xfs/xfs_attr_list.c b/fs/xfs/xfs_attr_list.c
>> index d37743b..881b9a4 100644
>> --- a/fs/xfs/xfs_attr_list.c
>> +++ b/fs/xfs/xfs_attr_list.c
>> @@ -12,6 +12,7 @@
>>   #include "xfs_trans_resv.h"
>>   #include "xfs_mount.h"
>>   #include "xfs_da_format.h"
>> +#include "xfs_da_btree.h"
>>   #include "xfs_inode.h"
>>   #include "xfs_trans.h"
>>   #include "xfs_bmap.h"
>> diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
>> index 28c07c9..7c1d9da 100644
>> --- a/fs/xfs/xfs_ioctl.c
>> +++ b/fs/xfs/xfs_ioctl.c
>> @@ -15,6 +15,8 @@
>>   #include "xfs_iwalk.h"
>>   #include "xfs_itable.h"
>>   #include "xfs_error.h"
>> +#include "xfs_da_format.h"
>> +#include "xfs_da_btree.h"
>>   #include "xfs_attr.h"
>>   #include "xfs_bmap.h"
>>   #include "xfs_bmap_util.h"
>> diff --git a/fs/xfs/xfs_ioctl32.c b/fs/xfs/xfs_ioctl32.c
>> index 769581a..d504f8f 100644
>> --- a/fs/xfs/xfs_ioctl32.c
>> +++ b/fs/xfs/xfs_ioctl32.c
>> @@ -17,6 +17,8 @@
>>   #include "xfs_itable.h"
>>   #include "xfs_fsops.h"
>>   #include "xfs_rtalloc.h"
>> +#include "xfs_da_format.h"
>> +#include "xfs_da_btree.h"
>>   #include "xfs_attr.h"
>>   #include "xfs_ioctl.h"
>>   #include "xfs_ioctl32.h"
>> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
>> index e85bbf5..a2d299f 100644
>> --- a/fs/xfs/xfs_iops.c
>> +++ b/fs/xfs/xfs_iops.c
>> @@ -13,6 +13,8 @@
>>   #include "xfs_inode.h"
>>   #include "xfs_acl.h"
>>   #include "xfs_quota.h"
>> +#include "xfs_da_format.h"
>> +#include "xfs_da_btree.h"
>>   #include "xfs_attr.h"
>>   #include "xfs_trans.h"
>>   #include "xfs_trace.h"
>> diff --git a/fs/xfs/xfs_xattr.c b/fs/xfs/xfs_xattr.c
>> index 74133a5..d8dc72d 100644
>> --- a/fs/xfs/xfs_xattr.c
>> +++ b/fs/xfs/xfs_xattr.c
>> @@ -10,6 +10,7 @@
>>   #include "xfs_log_format.h"
>>   #include "xfs_da_format.h"
>>   #include "xfs_inode.h"
>> +#include "xfs_da_btree.h"
>>   #include "xfs_attr.h"
>>   #include "xfs_acl.h"
>>   
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 16/16] xfs: Add delay ready attr set routines

[Allison Collins]

On 1/21/20 5:12 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:35PM -0700, Allison Collins wrote:
>> This patch modifies the attr set routines to be delay ready. This means
>> they no longer roll or commit transactions, but instead return -EAGAIN
>> to have the calling routine roll and refresh the transaction.  In this
>> series, xfs_attr_set_args has become xfs_attr_set_iter, which uses a
>> state machine like switch to keep track of where it was when EAGAIN was
>> returned.
>>
>> Part of xfs_attr_leaf_addname has been factored out into a new helper
>> function xfs_attr_leaf_try_add to allow transaction cycling between the
>> two routines.
>>
>> Two new helper functions have been added: xfs_attr_rmtval_set_init and
>> xfs_attr_rmtval_set_blk.  They provide a subset of logic similar to
>> xfs_attr_rmtval_set, but they store the current block in the delay attr
>> context to allow the caller to roll the transaction between allocations.
>> This helps to simplify and consolidate code used by
>> xfs_attr_leaf_addname and xfs_attr_node_addname
>>
>> Finally, xfs_attr_set_args has become a simple loop to refresh the
>> transaction until the operation is completed.
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c        | 347 +++++++++++++++++++++++++---------------
>>   fs/xfs/libxfs/xfs_attr.h        |   1 +
>>   fs/xfs/libxfs/xfs_attr_remote.c |  67 +++++++-
>>   fs/xfs/libxfs/xfs_attr_remote.h |   4 +
>>   fs/xfs/libxfs/xfs_da_btree.h    |  13 ++
>>   5 files changed, 301 insertions(+), 131 deletions(-)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index 92929ad..02a7f55 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -58,6 +58,7 @@ STATIC int xfs_attr_node_hasname(xfs_da_args_t *args,
>>   				 struct xfs_da_state **state);
>>   STATIC int xfs_attr_fillstate(xfs_da_state_t *state);
>>   STATIC int xfs_attr_refillstate(xfs_da_state_t *state);
>> +STATIC int xfs_attr_leaf_try_add(struct xfs_da_args *args, struct xfs_buf *bp);
>>   
>>   
>>   STATIC int
>> @@ -259,9 +260,86 @@ int
>>   xfs_attr_set_args(
>>   	struct xfs_da_args	*args)
>>   {
>> +	int			error = 0;
>> +	int			err2 = 0;
>> +	struct xfs_buf		*leaf_bp = NULL;
>> +
>> +	do {
>> +		error = xfs_attr_set_iter(args, &leaf_bp);
>> +		if (error && error != -EAGAIN)
>> +			goto out;
>> +
>> +		if (args->dac.flags & XFS_DAC_FINISH_TRANS) {
>> +			args->dac.flags &= ~XFS_DAC_FINISH_TRANS;
>> +
>> +			err2 = xfs_defer_finish(&args->trans);
>> +			if (err2) {
>> +				error = err2;
>> +				goto out;
>> +			}
>> +		}
>> +
>> +		err2 = xfs_trans_roll_inode(&args->trans, args->dp);
>> +		if (err2) {
>> +			error = err2;
>> +			goto out;
>> +		}
>> +
>> +		if (leaf_bp) {
>> +			xfs_trans_bjoin(args->trans, leaf_bp);
>> +			xfs_trans_bhold(args->trans, leaf_bp);
>> +		}
>> +
>> +	} while (error == -EAGAIN);
>> +
>> +out:
>> +	return error;
>> +}
>> +
>> +/*
>> + * Set the attribute specified in @args.
>> + * This routine is meant to function as a delayed operation, and may return
>> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
>> + * to handle this, and recall the function until a successful error code is
>> + * returned.
>> + */
>> +int
>> +xfs_attr_set_iter(
>> +	struct xfs_da_args	*args,
>> +	struct xfs_buf          **leaf_bp)
>> +{
>>   	struct xfs_inode	*dp = args->dp;
>> -	struct xfs_buf          *leaf_bp = NULL;
>> -	int			error, error2 = 0;
>> +	int			error = 0;
>> +	int			sf_size;
>> +
>> +	/* State machine switch */
>> +	switch (args->dac.dela_state) {
>> +	case XFS_DAS_SF_TO_LEAF:
>> +		goto sf_to_leaf;
>> +	case XFS_DAS_ALLOC_LEAF:
>> +	case XFS_DAS_FLIP_LFLAG:
>> +	case XFS_DAS_FOUND_LBLK:
>> +		goto leaf;
>> +	case XFS_DAS_FOUND_NBLK:
>> +	case XFS_DAS_FLIP_NFLAG:
>> +	case XFS_DAS_ALLOC_NODE:
>> +	case XFS_DAS_LEAF_TO_NODE:
>> +		goto node;
>> +	default:
> 
> Er... do these need to ASSERT on invalid states?  e.g. if somehow we end
> up with XFS_DAS_RM_SHRINK, we ought to blow up loudly, right?

We talked a little about validation checking in one of the previous reviews

https://www.spinics.net/lists/linux-xfs/msg34421.html

Ideally if we were going to do something like that, we should consider 
another layer of abstraction where we enforce the sequence of the states 
through some sort of "set_state()" function that checks for this. 
Though as mentioned in the thread, that's also a lot of extra over head.

otoh, at the time we discussed that, it wasnt very clear just what 
things were going to end up looking like.  So maybe at this point it's 
safe to hand wave it and say that states belonging to xfs_attr_set_iter 
should clearly never show up in xfs_attr_remove_iter or vis-a-vis.
> 
> 
>> +		break;
>> +	}
>> +
>> +	/*
>> +	 * New inodes may not have an attribute fork yet. So set the attribute
>> +	 * fork appropriately
>> +	 */
>> +	if (XFS_IFORK_Q((args->dp)) == 0) {
>> +		sf_size = sizeof(struct xfs_attr_sf_hdr) +
>> +		     XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
>> +		xfs_bmap_set_attrforkoff(args->dp, sf_size, NULL);
>> +		args->dp->i_afp = kmem_zone_zalloc(xfs_ifork_zone, 0);
>> +		args->dp->i_afp->if_flags = XFS_IFEXTENTS;
>> +	}
>>   
>>   	/*
>>   	 * If the attribute list is non-existent or a shortform list, proceed to
>> @@ -272,22 +350,20 @@ xfs_attr_set_args(
>>   	    (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
>>   	     dp->i_d.di_anextents == 0)))
>>   		goto sf_to_leaf;
>> -
>>   	/*
>>   	 * Try to add the attr to the attribute list in the inode.
>>   	 */
>>   	error = xfs_attr_try_sf_addname(dp, args);
>> -	if (error != -ENOSPC) {
>> -		error2 = xfs_trans_commit(args->trans);
>> -		args->trans = NULL;
>> -		return error ? error : error2;
> 
> Where does this commit go?
In this case, the add name is successful so the routine exists without 
needing to set a state.  The calling function (in this case 
xfs_attr_set) will finish out the commit.

> 
>> -	}
>> +
>> +	/* Should only be 0, -EEXIST or ENOSPC */
>> +	if (error != -ENOSPC)
>> +		return error;
>>   
>>   	/*
>>   	 * It won't fit in the shortform, transform to a leaf block.
>>   	 * GROT: another possible req'mt for a double-split btree op.
>>   	 */
>> -	error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
>> +	error = xfs_attr_shortform_to_leaf(args, leaf_bp);
>>   	if (error)
>>   		return error;
>>   
>> @@ -295,42 +371,47 @@ xfs_attr_set_args(
>>   	 * Prevent the leaf buffer from being unlocked so that a
>>   	 * concurrent AIL push cannot grab the half-baked leaf
>>   	 * buffer and run into problems with the write verifier.
>> -	 * Once we're done rolling the transaction we can release
>> -	 * the hold and add the attr to the leaf.
>>   	 */
>> -	xfs_trans_bhold(args->trans, leaf_bp);
>> -	error = xfs_defer_finish(&args->trans);
>> -	xfs_trans_bhold_release(args->trans, leaf_bp);
>> -	if (error) {
>> -		xfs_trans_brelse(args->trans, leaf_bp);
>> -		return error;
>> -	}
>> +	xfs_trans_bhold(args->trans, *leaf_bp);
>> +	args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +	args->dac.dela_state = XFS_DAS_SF_TO_LEAF;
>> +	return -EAGAIN;
>>   
>>   sf_to_leaf:
>>   
>> -	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>> -		error = xfs_attr_leaf_addname(args);
>> -		if (error != -ENOSPC)
>> -			return error;
>> -
>> -		/*
>> -		 * Commit that transaction so that the node_addname()
>> -		 * call can manage its own transactions.
>> -		 */
>> -		error = xfs_defer_finish(&args->trans);
>> -		if (error)
>> -			return error;
>> +	/*
>> +	 * After a shortform to leaf conversion, we need to hold the leaf and
>> +	 * cylce out the transaction.  When we get back, we need to release
>> +	 * the leaf.
>> +	 */
>> +	if (*leaf_bp != NULL) {
>> +		xfs_trans_brelse(args->trans, *leaf_bp);
>> +		*leaf_bp = NULL;
>> +	}
>>   
>> -		/*
>> -		 * Commit the current trans (including the inode) and
>> -		 * start a new one.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, dp);
>> -		if (error)
>> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>> +		error = xfs_attr_leaf_try_add(args, *leaf_bp);
>> +		switch (error) {
>> +		case -ENOSPC:
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +			args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>> +			return -EAGAIN;
>> +		case 0:
>> +			args->dac.dela_state = XFS_DAS_FOUND_LBLK;
>> +			return -EAGAIN;
>> +		default:
>>   			return error;
>> -
>> +		}
>> +leaf:
>> +		error = xfs_attr_leaf_addname(args);
>> +		if (error == -ENOSPC) {
>> +			args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>> +			return -EAGAIN;
>> +		}
>> +		return error;
>>   	}
>> -
>> +	args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>> +node:
>>   	error = xfs_attr_node_addname(args);
>>   	return error;
>>   }
>> @@ -759,27 +840,28 @@ xfs_attr_leaf_try_add(
>>    *
>>    * This leaf block cannot have a "remote" value, we only call this routine
>>    * if bmap_one_block() says there is only one block (ie: no remote blks).
>> + *
>> + * This routine is meant to function as a delayed operation, and may return
>> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
>> + * to handle this, and recall the function until a successful error code is
>> + * returned.
>>    */
>>   STATIC int
>>   xfs_attr_leaf_addname(struct xfs_da_args	*args)
>>   {
>> -	int			error, forkoff;
>>   	struct xfs_buf		*bp = NULL;
>> +	int			error, forkoff;
>>   	struct xfs_inode	*dp = args->dp;
>>   
>> -	trace_xfs_attr_leaf_addname(args);
>> -
>> -	error = xfs_attr_leaf_try_add(args, bp);
>> -	if (error)
>> -		return error;
>> -
>> -	/*
>> -	 * Commit the transaction that added the attr name so that
>> -	 * later routines can manage their own transactions.
>> -	 */
>> -	error = xfs_trans_roll_inode(&args->trans, dp);
>> -	if (error)
>> -		return error;
>> +	/* State machine switch */
>> +	switch (args->dac.dela_state) {
>> +	case XFS_DAS_FLIP_LFLAG:
>> +		goto flip_flag;
>> +	case XFS_DAS_ALLOC_LEAF:
>> +		goto alloc_leaf;
>> +	default:
>> +		break;
>> +	}
>>   
>>   	/*
>>   	 * If there was an out-of-line value, allocate the blocks we
>> @@ -788,7 +870,28 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>>   	 * maximum size of a transaction and/or hit a deadlock.
>>   	 */
>>   	if (args->rmtblkno > 0) {
>> -		error = xfs_attr_rmtval_set(args);
>> +
>> +		/* Open coded xfs_attr_rmtval_set without trans handling */
>> +		error = xfs_attr_rmtval_set_init(args);
>> +		if (error)
>> +			return error;
>> +
>> +		/*
>> +		 * Roll through the "value", allocating blocks on disk as
>> +		 * required.
>> +		 */
>> +alloc_leaf:
>> +		while (args->dac.blkcnt > 0) {
>> +			error = xfs_attr_rmtval_set_blk(args);
>> +			if (error)
>> +				return error;
>> +
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +			args->dac.dela_state = XFS_DAS_ALLOC_LEAF;
>> +			return -EAGAIN;
>> +		}
>> +
>> +		error = xfs_attr_rmtval_set_value(args);
>>   		if (error)
>>   			return error;
>>   	}
>> @@ -807,13 +910,10 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>>   		error = xfs_attr3_leaf_flipflags(args);
>>   		if (error)
>>   			return error;
>> -		/*
>> -		 * Commit the flag value change and start the next trans in
>> -		 * series.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>> -		if (error)
>> -			return error;
>> +
>> +		args->dac.dela_state = XFS_DAS_FLIP_LFLAG;
>> +		return -EAGAIN;
>> +flip_flag:
>>   
>>   		/*
>>   		 * Dismantle the "old" attribute/value pair by removing
>> @@ -844,34 +944,17 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>>   		/*
>>   		 * If the result is small enough, shrink it all into the inode.
>>   		 */
>> -		if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
>> +		forkoff = xfs_attr_shortform_allfit(bp, dp);
>> +		if (forkoff)
>>   			error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
>> -			/* bp is gone due to xfs_da_shrink_inode */
>> -			if (error)
>> -				return error;
>> -			error = xfs_defer_finish(&args->trans);
>> -			if (error)
>> -				return error;
>> -		}
>>   
>> -		/*
>> -		 * Commit the remove and start the next trans in series.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, dp);
>> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>   
>>   	} else if (args->rmtblkno > 0) {
>>   		/*
>>   		 * Added a "remote" value, just clear the incomplete flag.
>>   		 */
>>   		error = xfs_attr3_leaf_clearflag(args);
>> -		if (error)
>> -			return error;
>> -
>> -		/*
>> -		 * Commit the flag value change and start the next trans in
>> -		 * series.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>>   	}
>>   	return error;
>>   }
>> @@ -1020,16 +1103,22 @@ xfs_attr_node_hasname(
>>    *
>>    * "Remote" attribute values confuse the issue and atomic rename operations
>>    * add a whole extra layer of confusion on top of that.
>> + *
>> + * This routine is meant to function as a delayed operation, and may return
>> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
>> + * to handle this, and recall the function until a successful error code is
>> + *returned.
>>    */
>>   STATIC int
>>   xfs_attr_node_addname(
>>   	struct xfs_da_args	*args)
>>   {
>> -	struct xfs_da_state	*state;
>> +	struct xfs_da_state	*state = NULL;
>>   	struct xfs_da_state_blk	*blk;
>>   	struct xfs_inode	*dp;
>>   	struct xfs_mount	*mp;
>> -	int			retval, error;
>> +	int			retval = 0;
>> +	int			error = 0;
>>   
>>   	trace_xfs_attr_node_addname(args);
>>   
>> @@ -1038,7 +1127,19 @@ xfs_attr_node_addname(
>>   	 */
>>   	dp = args->dp;
>>   	mp = dp->i_mount;
>> -restart:
>> +
>> +	/* State machine switch */
>> +	switch (args->dac.dela_state) {
>> +	case XFS_DAS_FLIP_NFLAG:
>> +		goto flip_flag;
>> +	case XFS_DAS_FOUND_NBLK:
>> +		goto found_nblk;
>> +	case XFS_DAS_ALLOC_NODE:
>> +		goto alloc_node;
>> +	default:
>> +		break;
>> +	}
>> +
>>   	/*
>>   	 * Search to see if name already exists, and get back a pointer
>>   	 * to where it should go.
>> @@ -1088,19 +1189,13 @@ xfs_attr_node_addname(
>>   			error = xfs_attr3_leaf_to_node(args);
>>   			if (error)
>>   				goto out;
>> -			error = xfs_defer_finish(&args->trans);
>> -			if (error)
>> -				goto out;
>>   
>>   			/*
>> -			 * Commit the node conversion and start the next
>> -			 * trans in the chain.
>> +			 * Restart routine from the top.  No need to set  the
>> +			 * state
> 
> *OH*, dela_state == 0 does not mean "do not use delayed attrs", we're
> switching /all/ users to delayed attrs.  Urk.
> 
>>   			 */
>> -			error = xfs_trans_roll_inode(&args->trans, dp);
>> -			if (error)
>> -				goto out;
>> -
>> -			goto restart;
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +			return -EAGAIN;
>>   		}
>>   
>>   		/*
>> @@ -1112,9 +1207,7 @@ xfs_attr_node_addname(
>>   		error = xfs_da3_split(state);
>>   		if (error)
>>   			goto out;
>> -		error = xfs_defer_finish(&args->trans);
>> -		if (error)
>> -			goto out;
>> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>   	} else {
>>   		/*
>>   		 * Addition succeeded, update Btree hashvals.
>> @@ -1129,13 +1222,9 @@ xfs_attr_node_addname(
>>   	xfs_da_state_free(state);
>>   	state = NULL;
>>   
>> -	/*
>> -	 * Commit the leaf addition or btree split and start the next
>> -	 * trans in the chain.
>> -	 */
>> -	error = xfs_trans_roll_inode(&args->trans, dp);
>> -	if (error)
>> -		goto out;
>> +	args->dac.dela_state = XFS_DAS_FOUND_NBLK;
>> +	return -EAGAIN;
>> +found_nblk:
>>   
>>   	/*
>>   	 * If there was an out-of-line value, allocate the blocks we
>> @@ -1144,7 +1233,27 @@ xfs_attr_node_addname(
>>   	 * maximum size of a transaction and/or hit a deadlock.
>>   	 */
>>   	if (args->rmtblkno > 0) {
>> -		error = xfs_attr_rmtval_set(args);
>> +		/* Open coded xfs_attr_rmtval_set without trans handling */
>> +		error = xfs_attr_rmtval_set_init(args);
>> +		if (error)
>> +			return error;
>> +
>> +		/*
>> +		 * Roll through the "value", allocating blocks on disk as
>> +		 * required.
>> +		 */
>> +alloc_node:
>> +		while (args->dac.blkcnt > 0) {
>> +			error = xfs_attr_rmtval_set_blk(args);
>> +			if (error)
>> +				return error;
>> +
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +			args->dac.dela_state = XFS_DAS_ALLOC_NODE;
>> +			return -EAGAIN;
>> +		}
>> +
>> +		error = xfs_attr_rmtval_set_value(args);
>>   		if (error)
>>   			return error;
>>   	}
>> @@ -1167,10 +1276,9 @@ xfs_attr_node_addname(
>>   		 * Commit the flag value change and start the next trans in
>>   		 * series
>>   		 */
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>> -		if (error)
>> -			goto out;
>> -
>> +		args->dac.dela_state = XFS_DAS_FLIP_NFLAG;
>> +		return -EAGAIN;
>> +flip_flag:
>>   		/*
>>   		 * Dismantle the "old" attribute/value pair by removing
>>   		 * a "remote" value (if it exists).
>> @@ -1199,7 +1307,6 @@ xfs_attr_node_addname(
>>   		error = xfs_da3_node_lookup_int(state, &retval);
>>   		if (error)
>>   			goto out;
>> -
>>   		/*
>>   		 * Remove the name and update the hashvals in the tree.
>>   		 */
>> @@ -1207,7 +1314,6 @@ xfs_attr_node_addname(
>>   		ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
>>   		error = xfs_attr3_leaf_remove(blk->bp, args);
>>   		xfs_da3_fixhashpath(state, &state->path);
>> -
>>   		/*
>>   		 * Check to see if the tree needs to be collapsed.
>>   		 */
>> @@ -1215,18 +1321,9 @@ xfs_attr_node_addname(
>>   			error = xfs_da3_join(state);
>>   			if (error)
>>   				goto out;
>> -			error = xfs_defer_finish(&args->trans);
>> -			if (error)
>> -				goto out;
>> -		}
>> -
>> -		/*
>> -		 * Commit and start the next trans in the chain.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, dp);
>> -		if (error)
>> -			goto out;
>>   
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +		}
>>   	} else if (args->rmtblkno > 0) {
>>   		/*
>>   		 * Added a "remote" value, just clear the incomplete flag.
>> @@ -1234,14 +1331,6 @@ xfs_attr_node_addname(
>>   		error = xfs_attr3_leaf_clearflag(args);
>>   		if (error)
>>   			goto out;
>> -
>> -		 /*
>> -		  * Commit the flag value change and start the next trans in
>> -		  * series.
>> -		  */
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>> -		if (error)
>> -			goto out;
>>   	}
>>   	retval = error = 0;
>>   
>> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
>> index ea873a5..f450d8c 100644
>> --- a/fs/xfs/libxfs/xfs_attr.h
>> +++ b/fs/xfs/libxfs/xfs_attr.h
>> @@ -152,6 +152,7 @@ int xfs_attr_get(struct xfs_inode *ip, struct xfs_name *name,
>>   int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
>>   		 unsigned char *value, int valuelen, int flags);
>>   int xfs_attr_set_args(struct xfs_da_args *args);
>> +int xfs_attr_set_iter(struct xfs_da_args *args, struct xfs_buf **leaf_bp);
>>   int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
>>   int xfs_has_attr(struct xfs_da_args *args);
>>   int xfs_attr_remove_args(struct xfs_da_args *args);
>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
>> index f2ee0b8..ad4f14a 100644
>> --- a/fs/xfs/libxfs/xfs_attr_remote.c
>> +++ b/fs/xfs/libxfs/xfs_attr_remote.c
>> @@ -428,7 +428,7 @@ xfs_attr_rmtval_get(
>>    * Find a "hole" in the attribute address space large enough for us to drop the
>>    * new attribute's value into
>>    */
>> -STATIC int
>> +int
>>   xfs_attr_rmt_find_hole(
>>   	struct xfs_da_args	*args)
>>   {
>> @@ -455,7 +455,7 @@ xfs_attr_rmt_find_hole(
>>   	return 0;
>>   }
>>   
>> -STATIC int
>> +int
>>   xfs_attr_rmtval_set_value(
>>   	struct xfs_da_args	*args)
>>   {
>> @@ -588,6 +588,69 @@ xfs_attr_rmtval_set(
>>   }
>>   
>>   /*
>> + * Find a hole for the attr and store it in the delayed attr context.  This
>> + * initializes the context to roll through allocating an attr extent for a
>> + * delayed attr operation
>> + */
>> +int
>> +xfs_attr_rmtval_set_init(
>> +	struct xfs_da_args	*args)
>> +{
>> +	struct xfs_bmbt_irec	*map = &args->dac.map;
>> +	int error;
>> +
>> +	args->dac.lblkno = 0;
>> +	args->dac.lfileoff = 0;
>> +	args->dac.blkcnt = 0;
>> +	args->rmtblkcnt = 0;
>> +	args->rmtblkno = 0;
>> +	memset(map, 0, sizeof(struct xfs_bmbt_irec));
>> +
>> +	error = xfs_attr_rmt_find_hole(args);
>> +	if (error)
>> +		return error;
>> +
>> +	args->dac.blkcnt = args->rmtblkcnt;
>> +	args->dac.lblkno = args->rmtblkno;
>> +
>> +	return error;
>> +}
>> +
>> +/*
>> + * Write one block of the value associated with an attribute into the
>> + * out-of-line buffer that we have defined for it. This is similar to a subset
>> + * of xfs_attr_rmtval_set, but records the current block to the delayed attr
>> + * context, and leaves transaction handling to the caller.
>> + */
>> +int
>> +xfs_attr_rmtval_set_blk(
>> +	struct xfs_da_args	*args)
>> +{
>> +	struct xfs_inode	*dp = args->dp;
>> +	struct xfs_bmbt_irec	*map = &args->dac.map;
>> +	int nmap;
>> +	int error;
>> +
>> +	nmap = 1;
>> +	error = xfs_bmapi_write(args->trans, dp,
>> +		  (xfs_fileoff_t)args->dac.lblkno,
>> +		  args->dac.blkcnt, XFS_BMAPI_ATTRFORK,
>> +		  args->total, map, &nmap);
>> +	if (error)
>> +		return error;
>> +
>> +	ASSERT(nmap == 1);
>> +	ASSERT((map->br_startblock != DELAYSTARTBLOCK) &&
>> +	       (map->br_startblock != HOLESTARTBLOCK));
>> +
>> +	/* roll attribute extent map forwards */
>> +	args->dac.lblkno += map->br_blockcount;
>> +	args->dac.blkcnt -= map->br_blockcount;
>> +
>> +	return 0;
>> +}
>> +
>> +/*
>>    * Remove the value associated with an attribute by deleting the
>>    * out-of-line buffer that it is stored on.
>>    */
>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.h b/fs/xfs/libxfs/xfs_attr_remote.h
>> index 7ab3770..ab03519 100644
>> --- a/fs/xfs/libxfs/xfs_attr_remote.h
>> +++ b/fs/xfs/libxfs/xfs_attr_remote.h
>> @@ -13,4 +13,8 @@ int xfs_attr_rmtval_set(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_remove(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
>> +int xfs_attr_rmt_find_hole(struct xfs_da_args *args);
>> +int xfs_attr_rmtval_set_value(struct xfs_da_args *args);
>> +int xfs_attr_rmtval_set_blk(struct xfs_da_args *args);
>> +int xfs_attr_rmtval_set_init(struct xfs_da_args *args);
>>   #endif /* __XFS_ATTR_REMOTE_H__ */
>> diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
>> index 7fc87da..9943062 100644
>> --- a/fs/xfs/libxfs/xfs_da_btree.h
>> +++ b/fs/xfs/libxfs/xfs_da_btree.h
>> @@ -55,6 +55,14 @@ enum xfs_dacmp {
>>   enum xfs_delattr_state {
>>   	XFS_DAS_RM_SHRINK	= 1, /* We are shrinking the tree */
>>   	XFS_DAS_RM_NODE_BLKS	= 2, /* We are removing node blocks */
>> +	XFS_DAS_SF_TO_LEAF	= 3, /* Converted short form to leaf */
>> +	XFS_DAS_FOUND_LBLK	= 4, /* We found leaf blk for attr */
>> +	XFS_DAS_LEAF_TO_NODE	= 5, /* Converted leaf to node */
>> +	XFS_DAS_FOUND_NBLK	= 6, /* We found node blk for attr */
>> +	XFS_DAS_ALLOC_LEAF	= 7, /* We are allocating leaf blocks */
>> +	XFS_DAS_FLIP_LFLAG	= 8, /* Flipped leaf INCOMPLETE attr flag */
>> +	XFS_DAS_ALLOC_NODE	= 9, /* We are allocating node blocks */
>> +	XFS_DAS_FLIP_NFLAG	= 10,/* Flipped node INCOMPLETE attr flag */
> 
> We've definitely reached the point where a state diagram would be
> helpful.  Can you go from any of the RM_ states to the ones that you've
> just added?
No.  And if they did, they would have to show up in the calling 
functions state switches.  Because the calling function has to manage 
jumping straight back to the subroutine when ever the state belongs to 
the subroutine.  Kind of like how you see XFS_DAS_ALLOC_LEAF and 
XFS_DAS_FLIP_LFLAG in the state switch for xfs_attr_set_iter, even 
though those states only apply to xfs_attr_leaf_addname.

> 
> The new state machine code in last two patches would be a lot easier to
> review if I could look down from above instead of up from the XFS_DAS
> values and goto labels.  I /think/ it looks sane, but I'm only 20%
> confident of that statement.
> 
Sure, I'll see if I can put together a diagram to help folks out a bit.

Thanks for the reviews!
Allison

> --D
> 
>>   };
>>   
>>   /*
>> @@ -66,8 +74,13 @@ enum xfs_delattr_state {
>>    * Context used for keeping track of delayed attribute operations
>>    */
>>   struct xfs_delattr_context {
>> +	struct xfs_bmbt_irec	map;
>> +	struct xfs_buf		*leaf_bp;
>> +	xfs_fileoff_t		lfileoff;
>>   	struct xfs_da_state	*da_state;
>>   	struct xfs_da_state_blk *blk;
>> +	xfs_dablk_t		lblkno;
>> +	int			blkcnt;
>>   	int			flags;
>>   	enum xfs_delattr_state	dela_state;
>>   };
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 16/16] xfs: Add delay ready attr set routines

[Allison Collins]

On 1/21/20 5:12 PM, Darrick J. Wong wrote:
> On Sat, Jan 18, 2020 at 03:50:35PM -0700, Allison Collins wrote:
>> This patch modifies the attr set routines to be delay ready. This means
>> they no longer roll or commit transactions, but instead return -EAGAIN
>> to have the calling routine roll and refresh the transaction.  In this
>> series, xfs_attr_set_args has become xfs_attr_set_iter, which uses a
>> state machine like switch to keep track of where it was when EAGAIN was
>> returned.
>>
>> Part of xfs_attr_leaf_addname has been factored out into a new helper
>> function xfs_attr_leaf_try_add to allow transaction cycling between the
>> two routines.
>>
>> Two new helper functions have been added: xfs_attr_rmtval_set_init and
>> xfs_attr_rmtval_set_blk.  They provide a subset of logic similar to
>> xfs_attr_rmtval_set, but they store the current block in the delay attr
>> context to allow the caller to roll the transaction between allocations.
>> This helps to simplify and consolidate code used by
>> xfs_attr_leaf_addname and xfs_attr_node_addname
>>
>> Finally, xfs_attr_set_args has become a simple loop to refresh the
>> transaction until the operation is completed.
>>
>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>> ---
>>   fs/xfs/libxfs/xfs_attr.c        | 347 +++++++++++++++++++++++++---------------
>>   fs/xfs/libxfs/xfs_attr.h        |   1 +
>>   fs/xfs/libxfs/xfs_attr_remote.c |  67 +++++++-
>>   fs/xfs/libxfs/xfs_attr_remote.h |   4 +
>>   fs/xfs/libxfs/xfs_da_btree.h    |  13 ++
>>   5 files changed, 301 insertions(+), 131 deletions(-)
>>
>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>> index 92929ad..02a7f55 100644
>> --- a/fs/xfs/libxfs/xfs_attr.c
>> +++ b/fs/xfs/libxfs/xfs_attr.c
>> @@ -58,6 +58,7 @@ STATIC int xfs_attr_node_hasname(xfs_da_args_t *args,
>>   				 struct xfs_da_state **state);
>>   STATIC int xfs_attr_fillstate(xfs_da_state_t *state);
>>   STATIC int xfs_attr_refillstate(xfs_da_state_t *state);
>> +STATIC int xfs_attr_leaf_try_add(struct xfs_da_args *args, struct xfs_buf *bp);
>>   
>>   
>>   STATIC int
>> @@ -259,9 +260,86 @@ int
>>   xfs_attr_set_args(
>>   	struct xfs_da_args	*args)
>>   {
>> +	int			error = 0;
>> +	int			err2 = 0;
>> +	struct xfs_buf		*leaf_bp = NULL;
>> +
>> +	do {
>> +		error = xfs_attr_set_iter(args, &leaf_bp);
>> +		if (error && error != -EAGAIN)
>> +			goto out;
>> +
>> +		if (args->dac.flags & XFS_DAC_FINISH_TRANS) {
>> +			args->dac.flags &= ~XFS_DAC_FINISH_TRANS;
>> +
>> +			err2 = xfs_defer_finish(&args->trans);
>> +			if (err2) {
>> +				error = err2;
>> +				goto out;
>> +			}
>> +		}
>> +
>> +		err2 = xfs_trans_roll_inode(&args->trans, args->dp);
>> +		if (err2) {
>> +			error = err2;
>> +			goto out;
>> +		}
>> +
>> +		if (leaf_bp) {
>> +			xfs_trans_bjoin(args->trans, leaf_bp);
>> +			xfs_trans_bhold(args->trans, leaf_bp);
>> +		}
>> +
>> +	} while (error == -EAGAIN);
>> +
>> +out:
>> +	return error;
>> +}
>> +
>> +/*
>> + * Set the attribute specified in @args.
>> + * This routine is meant to function as a delayed operation, and may return
>> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
>> + * to handle this, and recall the function until a successful error code is
>> + * returned.
>> + */
>> +int
>> +xfs_attr_set_iter(
>> +	struct xfs_da_args	*args,
>> +	struct xfs_buf          **leaf_bp)
>> +{
>>   	struct xfs_inode	*dp = args->dp;
>> -	struct xfs_buf          *leaf_bp = NULL;
>> -	int			error, error2 = 0;
>> +	int			error = 0;
>> +	int			sf_size;
>> +
>> +	/* State machine switch */
>> +	switch (args->dac.dela_state) {
>> +	case XFS_DAS_SF_TO_LEAF:
>> +		goto sf_to_leaf;
>> +	case XFS_DAS_ALLOC_LEAF:
>> +	case XFS_DAS_FLIP_LFLAG:
>> +	case XFS_DAS_FOUND_LBLK:
>> +		goto leaf;
>> +	case XFS_DAS_FOUND_NBLK:
>> +	case XFS_DAS_FLIP_NFLAG:
>> +	case XFS_DAS_ALLOC_NODE:
>> +	case XFS_DAS_LEAF_TO_NODE:
>> +		goto node;
>> +	default:
> 
> Er... do these need to ASSERT on invalid states?  e.g. if somehow we end
> up with XFS_DAS_RM_SHRINK, we ought to blow up loudly, right?
> 
> 
>> +		break;
>> +	}
>> +
>> +	/*
>> +	 * New inodes may not have an attribute fork yet. So set the attribute
>> +	 * fork appropriately
>> +	 */
>> +	if (XFS_IFORK_Q((args->dp)) == 0) {
>> +		sf_size = sizeof(struct xfs_attr_sf_hdr) +
>> +		     XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, args->valuelen);
>> +		xfs_bmap_set_attrforkoff(args->dp, sf_size, NULL);
>> +		args->dp->i_afp = kmem_zone_zalloc(xfs_ifork_zone, 0);
>> +		args->dp->i_afp->if_flags = XFS_IFEXTENTS;
>> +	}
>>   
>>   	/*
>>   	 * If the attribute list is non-existent or a shortform list, proceed to
>> @@ -272,22 +350,20 @@ xfs_attr_set_args(
>>   	    (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
>>   	     dp->i_d.di_anextents == 0)))
>>   		goto sf_to_leaf;
>> -
>>   	/*
>>   	 * Try to add the attr to the attribute list in the inode.
>>   	 */
>>   	error = xfs_attr_try_sf_addname(dp, args);
>> -	if (error != -ENOSPC) {
>> -		error2 = xfs_trans_commit(args->trans);
>> -		args->trans = NULL;
>> -		return error ? error : error2;
> 
> Where does this commit go?
> 
>> -	}
>> +
>> +	/* Should only be 0, -EEXIST or ENOSPC */
>> +	if (error != -ENOSPC)
>> +		return error;
>>   
>>   	/*
>>   	 * It won't fit in the shortform, transform to a leaf block.
>>   	 * GROT: another possible req'mt for a double-split btree op.
>>   	 */
>> -	error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
>> +	error = xfs_attr_shortform_to_leaf(args, leaf_bp);
>>   	if (error)
>>   		return error;
>>   
>> @@ -295,42 +371,47 @@ xfs_attr_set_args(
>>   	 * Prevent the leaf buffer from being unlocked so that a
>>   	 * concurrent AIL push cannot grab the half-baked leaf
>>   	 * buffer and run into problems with the write verifier.
>> -	 * Once we're done rolling the transaction we can release
>> -	 * the hold and add the attr to the leaf.
>>   	 */
>> -	xfs_trans_bhold(args->trans, leaf_bp);
>> -	error = xfs_defer_finish(&args->trans);
>> -	xfs_trans_bhold_release(args->trans, leaf_bp);
>> -	if (error) {
>> -		xfs_trans_brelse(args->trans, leaf_bp);
>> -		return error;
>> -	}
>> +	xfs_trans_bhold(args->trans, *leaf_bp);
>> +	args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +	args->dac.dela_state = XFS_DAS_SF_TO_LEAF;
>> +	return -EAGAIN;
>>   
>>   sf_to_leaf:
>>   
>> -	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>> -		error = xfs_attr_leaf_addname(args);
>> -		if (error != -ENOSPC)
>> -			return error;
>> -
>> -		/*
>> -		 * Commit that transaction so that the node_addname()
>> -		 * call can manage its own transactions.
>> -		 */
>> -		error = xfs_defer_finish(&args->trans);
>> -		if (error)
>> -			return error;
>> +	/*
>> +	 * After a shortform to leaf conversion, we need to hold the leaf and
>> +	 * cylce out the transaction.  When we get back, we need to release
>> +	 * the leaf.
>> +	 */
>> +	if (*leaf_bp != NULL) {
>> +		xfs_trans_brelse(args->trans, *leaf_bp);
>> +		*leaf_bp = NULL;
>> +	}
>>   
>> -		/*
>> -		 * Commit the current trans (including the inode) and
>> -		 * start a new one.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, dp);
>> -		if (error)
>> +	if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>> +		error = xfs_attr_leaf_try_add(args, *leaf_bp);
>> +		switch (error) {
>> +		case -ENOSPC:
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +			args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>> +			return -EAGAIN;
>> +		case 0:
>> +			args->dac.dela_state = XFS_DAS_FOUND_LBLK;
>> +			return -EAGAIN;
>> +		default:
>>   			return error;
>> -
>> +		}
>> +leaf:
>> +		error = xfs_attr_leaf_addname(args);
>> +		if (error == -ENOSPC) {
>> +			args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>> +			return -EAGAIN;
>> +		}
>> +		return error;
>>   	}
>> -
>> +	args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>> +node:
>>   	error = xfs_attr_node_addname(args);
>>   	return error;
>>   }
>> @@ -759,27 +840,28 @@ xfs_attr_leaf_try_add(
>>    *
>>    * This leaf block cannot have a "remote" value, we only call this routine
>>    * if bmap_one_block() says there is only one block (ie: no remote blks).
>> + *
>> + * This routine is meant to function as a delayed operation, and may return
>> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
>> + * to handle this, and recall the function until a successful error code is
>> + * returned.
>>    */
>>   STATIC int
>>   xfs_attr_leaf_addname(struct xfs_da_args	*args)
>>   {
>> -	int			error, forkoff;
>>   	struct xfs_buf		*bp = NULL;
>> +	int			error, forkoff;
>>   	struct xfs_inode	*dp = args->dp;
>>   
>> -	trace_xfs_attr_leaf_addname(args);
>> -
>> -	error = xfs_attr_leaf_try_add(args, bp);
>> -	if (error)
>> -		return error;
>> -
>> -	/*
>> -	 * Commit the transaction that added the attr name so that
>> -	 * later routines can manage their own transactions.
>> -	 */
>> -	error = xfs_trans_roll_inode(&args->trans, dp);
>> -	if (error)
>> -		return error;
>> +	/* State machine switch */
>> +	switch (args->dac.dela_state) {
>> +	case XFS_DAS_FLIP_LFLAG:
>> +		goto flip_flag;
>> +	case XFS_DAS_ALLOC_LEAF:
>> +		goto alloc_leaf;
>> +	default:
>> +		break;
>> +	}
>>   
>>   	/*
>>   	 * If there was an out-of-line value, allocate the blocks we
>> @@ -788,7 +870,28 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>>   	 * maximum size of a transaction and/or hit a deadlock.
>>   	 */
>>   	if (args->rmtblkno > 0) {
>> -		error = xfs_attr_rmtval_set(args);
>> +
>> +		/* Open coded xfs_attr_rmtval_set without trans handling */
>> +		error = xfs_attr_rmtval_set_init(args);
>> +		if (error)
>> +			return error;
>> +
>> +		/*
>> +		 * Roll through the "value", allocating blocks on disk as
>> +		 * required.
>> +		 */
>> +alloc_leaf:
>> +		while (args->dac.blkcnt > 0) {
>> +			error = xfs_attr_rmtval_set_blk(args);
>> +			if (error)
>> +				return error;
>> +
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +			args->dac.dela_state = XFS_DAS_ALLOC_LEAF;
>> +			return -EAGAIN;
>> +		}
>> +
>> +		error = xfs_attr_rmtval_set_value(args);
>>   		if (error)
>>   			return error;
>>   	}
>> @@ -807,13 +910,10 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>>   		error = xfs_attr3_leaf_flipflags(args);
>>   		if (error)
>>   			return error;
>> -		/*
>> -		 * Commit the flag value change and start the next trans in
>> -		 * series.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>> -		if (error)
>> -			return error;
>> +
>> +		args->dac.dela_state = XFS_DAS_FLIP_LFLAG;
>> +		return -EAGAIN;
>> +flip_flag:
>>   
>>   		/*
>>   		 * Dismantle the "old" attribute/value pair by removing
>> @@ -844,34 +944,17 @@ xfs_attr_leaf_addname(struct xfs_da_args	*args)
>>   		/*
>>   		 * If the result is small enough, shrink it all into the inode.
>>   		 */
>> -		if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
>> +		forkoff = xfs_attr_shortform_allfit(bp, dp);
>> +		if (forkoff)
>>   			error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
>> -			/* bp is gone due to xfs_da_shrink_inode */
>> -			if (error)
>> -				return error;
>> -			error = xfs_defer_finish(&args->trans);
>> -			if (error)
>> -				return error;
>> -		}
>>   
>> -		/*
>> -		 * Commit the remove and start the next trans in series.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, dp);
>> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>   
>>   	} else if (args->rmtblkno > 0) {
>>   		/*
>>   		 * Added a "remote" value, just clear the incomplete flag.
>>   		 */
>>   		error = xfs_attr3_leaf_clearflag(args);
>> -		if (error)
>> -			return error;
>> -
>> -		/*
>> -		 * Commit the flag value change and start the next trans in
>> -		 * series.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>>   	}
>>   	return error;
>>   }
>> @@ -1020,16 +1103,22 @@ xfs_attr_node_hasname(
>>    *
>>    * "Remote" attribute values confuse the issue and atomic rename operations
>>    * add a whole extra layer of confusion on top of that.
>> + *
>> + * This routine is meant to function as a delayed operation, and may return
>> + * -EAGAIN when the transaction needs to be rolled.  Calling functions will need
>> + * to handle this, and recall the function until a successful error code is
>> + *returned.
>>    */
>>   STATIC int
>>   xfs_attr_node_addname(
>>   	struct xfs_da_args	*args)
>>   {
>> -	struct xfs_da_state	*state;
>> +	struct xfs_da_state	*state = NULL;
>>   	struct xfs_da_state_blk	*blk;
>>   	struct xfs_inode	*dp;
>>   	struct xfs_mount	*mp;
>> -	int			retval, error;
>> +	int			retval = 0;
>> +	int			error = 0;
>>   
>>   	trace_xfs_attr_node_addname(args);
>>   
>> @@ -1038,7 +1127,19 @@ xfs_attr_node_addname(
>>   	 */
>>   	dp = args->dp;
>>   	mp = dp->i_mount;
>> -restart:
>> +
>> +	/* State machine switch */
>> +	switch (args->dac.dela_state) {
>> +	case XFS_DAS_FLIP_NFLAG:
>> +		goto flip_flag;
>> +	case XFS_DAS_FOUND_NBLK:
>> +		goto found_nblk;
>> +	case XFS_DAS_ALLOC_NODE:
>> +		goto alloc_node;
>> +	default:
>> +		break;
>> +	}
>> +
>>   	/*
>>   	 * Search to see if name already exists, and get back a pointer
>>   	 * to where it should go.
>> @@ -1088,19 +1189,13 @@ xfs_attr_node_addname(
>>   			error = xfs_attr3_leaf_to_node(args);
>>   			if (error)
>>   				goto out;
>> -			error = xfs_defer_finish(&args->trans);
>> -			if (error)
>> -				goto out;
>>   
>>   			/*
>> -			 * Commit the node conversion and start the next
>> -			 * trans in the chain.
>> +			 * Restart routine from the top.  No need to set  the
>> +			 * state
> 

I think I missed responding to this comment in my last email:

> *OH*, dela_state == 0 does not mean "do not use delayed attrs", we're
> switching /all/ users to delayed attrs.  Urk.

Not quite.  While both inline and delayed operations go through this 
same dance of EAGAIN ping pong, it's not a delayed operation until we 
package it into an intent and log it.  And then the intent pops back out 
during defer finish, and the ping pong starts from 
xfs_defer_finish_noroll instead of xfs_attr_set_iter.

We still need to support in line operations for older versions that do 
not support the attr log item.  Later when we get into delayed attrs, we 
use a feature bit to control this, and toggle as needed.

Hope that helps!

> 
>>   			 */
>> -			error = xfs_trans_roll_inode(&args->trans, dp);
>> -			if (error)
>> -				goto out;
>> -
>> -			goto restart;
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +			return -EAGAIN;
>>   		}
>>   
>>   		/*
>> @@ -1112,9 +1207,7 @@ xfs_attr_node_addname(
>>   		error = xfs_da3_split(state);
>>   		if (error)
>>   			goto out;
>> -		error = xfs_defer_finish(&args->trans);
>> -		if (error)
>> -			goto out;
>> +		args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>   	} else {
>>   		/*
>>   		 * Addition succeeded, update Btree hashvals.
>> @@ -1129,13 +1222,9 @@ xfs_attr_node_addname(
>>   	xfs_da_state_free(state);
>>   	state = NULL;
>>   
>> -	/*
>> -	 * Commit the leaf addition or btree split and start the next
>> -	 * trans in the chain.
>> -	 */
>> -	error = xfs_trans_roll_inode(&args->trans, dp);
>> -	if (error)
>> -		goto out;
>> +	args->dac.dela_state = XFS_DAS_FOUND_NBLK;
>> +	return -EAGAIN;
>> +found_nblk:
>>   
>>   	/*
>>   	 * If there was an out-of-line value, allocate the blocks we
>> @@ -1144,7 +1233,27 @@ xfs_attr_node_addname(
>>   	 * maximum size of a transaction and/or hit a deadlock.
>>   	 */
>>   	if (args->rmtblkno > 0) {
>> -		error = xfs_attr_rmtval_set(args);
>> +		/* Open coded xfs_attr_rmtval_set without trans handling */
>> +		error = xfs_attr_rmtval_set_init(args);
>> +		if (error)
>> +			return error;
>> +
>> +		/*
>> +		 * Roll through the "value", allocating blocks on disk as
>> +		 * required.
>> +		 */
>> +alloc_node:
>> +		while (args->dac.blkcnt > 0) {
>> +			error = xfs_attr_rmtval_set_blk(args);
>> +			if (error)
>> +				return error;
>> +
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +			args->dac.dela_state = XFS_DAS_ALLOC_NODE;
>> +			return -EAGAIN;
>> +		}
>> +
>> +		error = xfs_attr_rmtval_set_value(args);
>>   		if (error)
>>   			return error;
>>   	}
>> @@ -1167,10 +1276,9 @@ xfs_attr_node_addname(
>>   		 * Commit the flag value change and start the next trans in
>>   		 * series
>>   		 */
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>> -		if (error)
>> -			goto out;
>> -
>> +		args->dac.dela_state = XFS_DAS_FLIP_NFLAG;
>> +		return -EAGAIN;
>> +flip_flag:
>>   		/*
>>   		 * Dismantle the "old" attribute/value pair by removing
>>   		 * a "remote" value (if it exists).
>> @@ -1199,7 +1307,6 @@ xfs_attr_node_addname(
>>   		error = xfs_da3_node_lookup_int(state, &retval);
>>   		if (error)
>>   			goto out;
>> -
>>   		/*
>>   		 * Remove the name and update the hashvals in the tree.
>>   		 */
>> @@ -1207,7 +1314,6 @@ xfs_attr_node_addname(
>>   		ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
>>   		error = xfs_attr3_leaf_remove(blk->bp, args);
>>   		xfs_da3_fixhashpath(state, &state->path);
>> -
>>   		/*
>>   		 * Check to see if the tree needs to be collapsed.
>>   		 */
>> @@ -1215,18 +1321,9 @@ xfs_attr_node_addname(
>>   			error = xfs_da3_join(state);
>>   			if (error)
>>   				goto out;
>> -			error = xfs_defer_finish(&args->trans);
>> -			if (error)
>> -				goto out;
>> -		}
>> -
>> -		/*
>> -		 * Commit and start the next trans in the chain.
>> -		 */
>> -		error = xfs_trans_roll_inode(&args->trans, dp);
>> -		if (error)
>> -			goto out;
>>   
>> +			args->dac.flags |= XFS_DAC_FINISH_TRANS;
>> +		}
>>   	} else if (args->rmtblkno > 0) {
>>   		/*
>>   		 * Added a "remote" value, just clear the incomplete flag.
>> @@ -1234,14 +1331,6 @@ xfs_attr_node_addname(
>>   		error = xfs_attr3_leaf_clearflag(args);
>>   		if (error)
>>   			goto out;
>> -
>> -		 /*
>> -		  * Commit the flag value change and start the next trans in
>> -		  * series.
>> -		  */
>> -		error = xfs_trans_roll_inode(&args->trans, args->dp);
>> -		if (error)
>> -			goto out;
>>   	}
>>   	retval = error = 0;
>>   
>> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
>> index ea873a5..f450d8c 100644
>> --- a/fs/xfs/libxfs/xfs_attr.h
>> +++ b/fs/xfs/libxfs/xfs_attr.h
>> @@ -152,6 +152,7 @@ int xfs_attr_get(struct xfs_inode *ip, struct xfs_name *name,
>>   int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
>>   		 unsigned char *value, int valuelen, int flags);
>>   int xfs_attr_set_args(struct xfs_da_args *args);
>> +int xfs_attr_set_iter(struct xfs_da_args *args, struct xfs_buf **leaf_bp);
>>   int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, int flags);
>>   int xfs_has_attr(struct xfs_da_args *args);
>>   int xfs_attr_remove_args(struct xfs_da_args *args);
>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
>> index f2ee0b8..ad4f14a 100644
>> --- a/fs/xfs/libxfs/xfs_attr_remote.c
>> +++ b/fs/xfs/libxfs/xfs_attr_remote.c
>> @@ -428,7 +428,7 @@ xfs_attr_rmtval_get(
>>    * Find a "hole" in the attribute address space large enough for us to drop the
>>    * new attribute's value into
>>    */
>> -STATIC int
>> +int
>>   xfs_attr_rmt_find_hole(
>>   	struct xfs_da_args	*args)
>>   {
>> @@ -455,7 +455,7 @@ xfs_attr_rmt_find_hole(
>>   	return 0;
>>   }
>>   
>> -STATIC int
>> +int
>>   xfs_attr_rmtval_set_value(
>>   	struct xfs_da_args	*args)
>>   {
>> @@ -588,6 +588,69 @@ xfs_attr_rmtval_set(
>>   }
>>   
>>   /*
>> + * Find a hole for the attr and store it in the delayed attr context.  This
>> + * initializes the context to roll through allocating an attr extent for a
>> + * delayed attr operation
>> + */
>> +int
>> +xfs_attr_rmtval_set_init(
>> +	struct xfs_da_args	*args)
>> +{
>> +	struct xfs_bmbt_irec	*map = &args->dac.map;
>> +	int error;
>> +
>> +	args->dac.lblkno = 0;
>> +	args->dac.lfileoff = 0;
>> +	args->dac.blkcnt = 0;
>> +	args->rmtblkcnt = 0;
>> +	args->rmtblkno = 0;
>> +	memset(map, 0, sizeof(struct xfs_bmbt_irec));
>> +
>> +	error = xfs_attr_rmt_find_hole(args);
>> +	if (error)
>> +		return error;
>> +
>> +	args->dac.blkcnt = args->rmtblkcnt;
>> +	args->dac.lblkno = args->rmtblkno;
>> +
>> +	return error;
>> +}
>> +
>> +/*
>> + * Write one block of the value associated with an attribute into the
>> + * out-of-line buffer that we have defined for it. This is similar to a subset
>> + * of xfs_attr_rmtval_set, but records the current block to the delayed attr
>> + * context, and leaves transaction handling to the caller.
>> + */
>> +int
>> +xfs_attr_rmtval_set_blk(
>> +	struct xfs_da_args	*args)
>> +{
>> +	struct xfs_inode	*dp = args->dp;
>> +	struct xfs_bmbt_irec	*map = &args->dac.map;
>> +	int nmap;
>> +	int error;
>> +
>> +	nmap = 1;
>> +	error = xfs_bmapi_write(args->trans, dp,
>> +		  (xfs_fileoff_t)args->dac.lblkno,
>> +		  args->dac.blkcnt, XFS_BMAPI_ATTRFORK,
>> +		  args->total, map, &nmap);
>> +	if (error)
>> +		return error;
>> +
>> +	ASSERT(nmap == 1);
>> +	ASSERT((map->br_startblock != DELAYSTARTBLOCK) &&
>> +	       (map->br_startblock != HOLESTARTBLOCK));
>> +
>> +	/* roll attribute extent map forwards */
>> +	args->dac.lblkno += map->br_blockcount;
>> +	args->dac.blkcnt -= map->br_blockcount;
>> +
>> +	return 0;
>> +}
>> +
>> +/*
>>    * Remove the value associated with an attribute by deleting the
>>    * out-of-line buffer that it is stored on.
>>    */
>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.h b/fs/xfs/libxfs/xfs_attr_remote.h
>> index 7ab3770..ab03519 100644
>> --- a/fs/xfs/libxfs/xfs_attr_remote.h
>> +++ b/fs/xfs/libxfs/xfs_attr_remote.h
>> @@ -13,4 +13,8 @@ int xfs_attr_rmtval_set(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_remove(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
>>   int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
>> +int xfs_attr_rmt_find_hole(struct xfs_da_args *args);
>> +int xfs_attr_rmtval_set_value(struct xfs_da_args *args);
>> +int xfs_attr_rmtval_set_blk(struct xfs_da_args *args);
>> +int xfs_attr_rmtval_set_init(struct xfs_da_args *args);
>>   #endif /* __XFS_ATTR_REMOTE_H__ */
>> diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
>> index 7fc87da..9943062 100644
>> --- a/fs/xfs/libxfs/xfs_da_btree.h
>> +++ b/fs/xfs/libxfs/xfs_da_btree.h
>> @@ -55,6 +55,14 @@ enum xfs_dacmp {
>>   enum xfs_delattr_state {
>>   	XFS_DAS_RM_SHRINK	= 1, /* We are shrinking the tree */
>>   	XFS_DAS_RM_NODE_BLKS	= 2, /* We are removing node blocks */
>> +	XFS_DAS_SF_TO_LEAF	= 3, /* Converted short form to leaf */
>> +	XFS_DAS_FOUND_LBLK	= 4, /* We found leaf blk for attr */
>> +	XFS_DAS_LEAF_TO_NODE	= 5, /* Converted leaf to node */
>> +	XFS_DAS_FOUND_NBLK	= 6, /* We found node blk for attr */
>> +	XFS_DAS_ALLOC_LEAF	= 7, /* We are allocating leaf blocks */
>> +	XFS_DAS_FLIP_LFLAG	= 8, /* Flipped leaf INCOMPLETE attr flag */
>> +	XFS_DAS_ALLOC_NODE	= 9, /* We are allocating node blocks */
>> +	XFS_DAS_FLIP_NFLAG	= 10,/* Flipped node INCOMPLETE attr flag */
> 
> We've definitely reached the point where a state diagram would be
> helpful.  Can you go from any of the RM_ states to the ones that you've
> just added?
> 
> The new state machine code in last two patches would be a lot easier to
> review if I could look down from above instead of up from the XFS_DAS
> values and goto labels.  I /think/ it looks sane, but I'm only 20%
> confident of that statement.
> 
> --D
> 
>>   };
>>   
>>   /*
>> @@ -66,8 +74,13 @@ enum xfs_delattr_state {
>>    * Context used for keeping track of delayed attribute operations
>>    */
>>   struct xfs_delattr_context {
>> +	struct xfs_bmbt_irec	map;
>> +	struct xfs_buf		*leaf_bp;
>> +	xfs_fileoff_t		lfileoff;
>>   	struct xfs_da_state	*da_state;
>>   	struct xfs_da_state_blk *blk;
>> +	xfs_dablk_t		lblkno;
>> +	int			blkcnt;
>>   	int			flags;
>>   	enum xfs_delattr_state	dela_state;
>>   };
>> -- 
>> 2.7.4
>>

Re: [PATCH v6 16/16] xfs: Add delay ready attr set routines

[Allison Collins]

On 1/22/20 9:57 AM, Allison Collins wrote:
> 
> 
> On 1/21/20 5:12 PM, Darrick J. Wong wrote:
>> On Sat, Jan 18, 2020 at 03:50:35PM -0700, Allison Collins wrote:
>>> This patch modifies the attr set routines to be delay ready. This means
>>> they no longer roll or commit transactions, but instead return -EAGAIN
>>> to have the calling routine roll and refresh the transaction.  In this
>>> series, xfs_attr_set_args has become xfs_attr_set_iter, which uses a
>>> state machine like switch to keep track of where it was when EAGAIN was
>>> returned.
>>>
>>> Part of xfs_attr_leaf_addname has been factored out into a new helper
>>> function xfs_attr_leaf_try_add to allow transaction cycling between the
>>> two routines.
>>>
>>> Two new helper functions have been added: xfs_attr_rmtval_set_init and
>>> xfs_attr_rmtval_set_blk.  They provide a subset of logic similar to
>>> xfs_attr_rmtval_set, but they store the current block in the delay attr
>>> context to allow the caller to roll the transaction between allocations.
>>> This helps to simplify and consolidate code used by
>>> xfs_attr_leaf_addname and xfs_attr_node_addname
>>>
>>> Finally, xfs_attr_set_args has become a simple loop to refresh the
>>> transaction until the operation is completed.
>>>
>>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>>> ---
>>>   fs/xfs/libxfs/xfs_attr.c        | 347 
>>> +++++++++++++++++++++++++---------------
>>>   fs/xfs/libxfs/xfs_attr.h        |   1 +
>>>   fs/xfs/libxfs/xfs_attr_remote.c |  67 +++++++-
>>>   fs/xfs/libxfs/xfs_attr_remote.h |   4 +
>>>   fs/xfs/libxfs/xfs_da_btree.h    |  13 ++
>>>   5 files changed, 301 insertions(+), 131 deletions(-)
>>>
>>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>>> index 92929ad..02a7f55 100644
>>> --- a/fs/xfs/libxfs/xfs_attr.c
>>> +++ b/fs/xfs/libxfs/xfs_attr.c
>>> @@ -58,6 +58,7 @@ STATIC int xfs_attr_node_hasname(xfs_da_args_t *args,
>>>                    struct xfs_da_state **state);
>>>   STATIC int xfs_attr_fillstate(xfs_da_state_t *state);
>>>   STATIC int xfs_attr_refillstate(xfs_da_state_t *state);
>>> +STATIC int xfs_attr_leaf_try_add(struct xfs_da_args *args, struct 
>>> xfs_buf *bp);
>>>   STATIC int
>>> @@ -259,9 +260,86 @@ int
>>>   xfs_attr_set_args(
>>>       struct xfs_da_args    *args)
>>>   {
>>> +    int            error = 0;
>>> +    int            err2 = 0;
>>> +    struct xfs_buf        *leaf_bp = NULL;
>>> +
>>> +    do {
>>> +        error = xfs_attr_set_iter(args, &leaf_bp);
>>> +        if (error && error != -EAGAIN)
>>> +            goto out;
>>> +
>>> +        if (args->dac.flags & XFS_DAC_FINISH_TRANS) {
>>> +            args->dac.flags &= ~XFS_DAC_FINISH_TRANS;
>>> +
>>> +            err2 = xfs_defer_finish(&args->trans);
>>> +            if (err2) {
>>> +                error = err2;
>>> +                goto out;
>>> +            }
>>> +        }
>>> +
>>> +        err2 = xfs_trans_roll_inode(&args->trans, args->dp);
>>> +        if (err2) {
>>> +            error = err2;
>>> +            goto out;
>>> +        }
>>> +
>>> +        if (leaf_bp) {
>>> +            xfs_trans_bjoin(args->trans, leaf_bp);
>>> +            xfs_trans_bhold(args->trans, leaf_bp);
>>> +        }
>>> +
>>> +    } while (error == -EAGAIN);
>>> +
>>> +out:
>>> +    return error;
>>> +}
>>> +
>>> +/*
>>> + * Set the attribute specified in @args.
>>> + * This routine is meant to function as a delayed operation, and may 
>>> return
>>> + * -EAGAIN when the transaction needs to be rolled.  Calling 
>>> functions will need
>>> + * to handle this, and recall the function until a successful error 
>>> code is
>>> + * returned.
>>> + */
>>> +int
>>> +xfs_attr_set_iter(
>>> +    struct xfs_da_args    *args,
>>> +    struct xfs_buf          **leaf_bp)
>>> +{
>>>       struct xfs_inode    *dp = args->dp;
>>> -    struct xfs_buf          *leaf_bp = NULL;
>>> -    int            error, error2 = 0;
>>> +    int            error = 0;
>>> +    int            sf_size;
>>> +
>>> +    /* State machine switch */
>>> +    switch (args->dac.dela_state) {
>>> +    case XFS_DAS_SF_TO_LEAF:
>>> +        goto sf_to_leaf;
>>> +    case XFS_DAS_ALLOC_LEAF:
>>> +    case XFS_DAS_FLIP_LFLAG:
>>> +    case XFS_DAS_FOUND_LBLK:
>>> +        goto leaf;
>>> +    case XFS_DAS_FOUND_NBLK:
>>> +    case XFS_DAS_FLIP_NFLAG:
>>> +    case XFS_DAS_ALLOC_NODE:
>>> +    case XFS_DAS_LEAF_TO_NODE:
>>> +        goto node;
>>> +    default:
>>
>> Er... do these need to ASSERT on invalid states?  e.g. if somehow we end
>> up with XFS_DAS_RM_SHRINK, we ought to blow up loudly, right?
>>
>>
>>> +        break;
>>> +    }
>>> +
>>> +    /*
>>> +     * New inodes may not have an attribute fork yet. So set the 
>>> attribute
>>> +     * fork appropriately
>>> +     */
>>> +    if (XFS_IFORK_Q((args->dp)) == 0) {
>>> +        sf_size = sizeof(struct xfs_attr_sf_hdr) +
>>> +             XFS_ATTR_SF_ENTSIZE_BYNAME(args->name.len, 
>>> args->valuelen);
>>> +        xfs_bmap_set_attrforkoff(args->dp, sf_size, NULL);
>>> +        args->dp->i_afp = kmem_zone_zalloc(xfs_ifork_zone, 0);
>>> +        args->dp->i_afp->if_flags = XFS_IFEXTENTS;
>>> +    }
>>>       /*
>>>        * If the attribute list is non-existent or a shortform list, 
>>> proceed to
>>> @@ -272,22 +350,20 @@ xfs_attr_set_args(
>>>           (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
>>>            dp->i_d.di_anextents == 0)))
>>>           goto sf_to_leaf;
>>> -
>>>       /*
>>>        * Try to add the attr to the attribute list in the inode.
>>>        */
>>>       error = xfs_attr_try_sf_addname(dp, args);
>>> -    if (error != -ENOSPC) {
>>> -        error2 = xfs_trans_commit(args->trans);
>>> -        args->trans = NULL;
>>> -        return error ? error : error2;
>>
>> Where does this commit go?
>>
>>> -    }
>>> +
>>> +    /* Should only be 0, -EEXIST or ENOSPC */
>>> +    if (error != -ENOSPC)
>>> +        return error;
>>>       /*
>>>        * It won't fit in the shortform, transform to a leaf block.
>>>        * GROT: another possible req'mt for a double-split btree op.
>>>        */
>>> -    error = xfs_attr_shortform_to_leaf(args, &leaf_bp);
>>> +    error = xfs_attr_shortform_to_leaf(args, leaf_bp);
>>>       if (error)
>>>           return error;
>>> @@ -295,42 +371,47 @@ xfs_attr_set_args(
>>>        * Prevent the leaf buffer from being unlocked so that a
>>>        * concurrent AIL push cannot grab the half-baked leaf
>>>        * buffer and run into problems with the write verifier.
>>> -     * Once we're done rolling the transaction we can release
>>> -     * the hold and add the attr to the leaf.
>>>        */
>>> -    xfs_trans_bhold(args->trans, leaf_bp);
>>> -    error = xfs_defer_finish(&args->trans);
>>> -    xfs_trans_bhold_release(args->trans, leaf_bp);
>>> -    if (error) {
>>> -        xfs_trans_brelse(args->trans, leaf_bp);
>>> -        return error;
>>> -    }
>>> +    xfs_trans_bhold(args->trans, *leaf_bp);
>>> +    args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>> +    args->dac.dela_state = XFS_DAS_SF_TO_LEAF;
>>> +    return -EAGAIN;
>>>   sf_to_leaf:
>>> -    if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>>> -        error = xfs_attr_leaf_addname(args);
>>> -        if (error != -ENOSPC)
>>> -            return error;
>>> -
>>> -        /*
>>> -         * Commit that transaction so that the node_addname()
>>> -         * call can manage its own transactions.
>>> -         */
>>> -        error = xfs_defer_finish(&args->trans);
>>> -        if (error)
>>> -            return error;
>>> +    /*
>>> +     * After a shortform to leaf conversion, we need to hold the 
>>> leaf and
>>> +     * cylce out the transaction.  When we get back, we need to release
>>> +     * the leaf.
>>> +     */
>>> +    if (*leaf_bp != NULL) {
>>> +        xfs_trans_brelse(args->trans, *leaf_bp);
>>> +        *leaf_bp = NULL;
>>> +    }
>>> -        /*
>>> -         * Commit the current trans (including the inode) and
>>> -         * start a new one.
>>> -         */
>>> -        error = xfs_trans_roll_inode(&args->trans, dp);
>>> -        if (error)
>>> +    if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) {
>>> +        error = xfs_attr_leaf_try_add(args, *leaf_bp);
>>> +        switch (error) {
>>> +        case -ENOSPC:
>>> +            args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>> +            args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>>> +            return -EAGAIN;
>>> +        case 0:
>>> +            args->dac.dela_state = XFS_DAS_FOUND_LBLK;
>>> +            return -EAGAIN;
>>> +        default:
>>>               return error;
>>> -
>>> +        }
>>> +leaf:
>>> +        error = xfs_attr_leaf_addname(args);
>>> +        if (error == -ENOSPC) {
>>> +            args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>>> +            return -EAGAIN;
>>> +        }
>>> +        return error;
>>>       }
>>> -
>>> +    args->dac.dela_state = XFS_DAS_LEAF_TO_NODE;
>>> +node:
>>>       error = xfs_attr_node_addname(args);
>>>       return error;
>>>   }
>>> @@ -759,27 +840,28 @@ xfs_attr_leaf_try_add(
>>>    *
>>>    * This leaf block cannot have a "remote" value, we only call this 
>>> routine
>>>    * if bmap_one_block() says there is only one block (ie: no remote 
>>> blks).
>>> + *
>>> + * This routine is meant to function as a delayed operation, and may 
>>> return
>>> + * -EAGAIN when the transaction needs to be rolled.  Calling 
>>> functions will need
>>> + * to handle this, and recall the function until a successful error 
>>> code is
>>> + * returned.
>>>    */
>>>   STATIC int
>>>   xfs_attr_leaf_addname(struct xfs_da_args    *args)
>>>   {
>>> -    int            error, forkoff;
>>>       struct xfs_buf        *bp = NULL;
>>> +    int            error, forkoff;
>>>       struct xfs_inode    *dp = args->dp;
>>> -    trace_xfs_attr_leaf_addname(args);
>>> -
>>> -    error = xfs_attr_leaf_try_add(args, bp);
>>> -    if (error)
>>> -        return error;
>>> -
>>> -    /*
>>> -     * Commit the transaction that added the attr name so that
>>> -     * later routines can manage their own transactions.
>>> -     */
>>> -    error = xfs_trans_roll_inode(&args->trans, dp);
>>> -    if (error)
>>> -        return error;
>>> +    /* State machine switch */
>>> +    switch (args->dac.dela_state) {
>>> +    case XFS_DAS_FLIP_LFLAG:
>>> +        goto flip_flag;
>>> +    case XFS_DAS_ALLOC_LEAF:
>>> +        goto alloc_leaf;
>>> +    default:
>>> +        break;
>>> +    }
>>>       /*
>>>        * If there was an out-of-line value, allocate the blocks we
>>> @@ -788,7 +870,28 @@ xfs_attr_leaf_addname(struct xfs_da_args    *args)
>>>        * maximum size of a transaction and/or hit a deadlock.
>>>        */
>>>       if (args->rmtblkno > 0) {
>>> -        error = xfs_attr_rmtval_set(args);
>>> +
>>> +        /* Open coded xfs_attr_rmtval_set without trans handling */
>>> +        error = xfs_attr_rmtval_set_init(args);
>>> +        if (error)
>>> +            return error;
>>> +
>>> +        /*
>>> +         * Roll through the "value", allocating blocks on disk as
>>> +         * required.
>>> +         */
>>> +alloc_leaf:
>>> +        while (args->dac.blkcnt > 0) {
>>> +            error = xfs_attr_rmtval_set_blk(args);
>>> +            if (error)
>>> +                return error;
>>> +
>>> +            args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>> +            args->dac.dela_state = XFS_DAS_ALLOC_LEAF;
>>> +            return -EAGAIN;
>>> +        }
>>> +
>>> +        error = xfs_attr_rmtval_set_value(args);
>>>           if (error)
>>>               return error;
>>>       }
>>> @@ -807,13 +910,10 @@ xfs_attr_leaf_addname(struct xfs_da_args    *args)
>>>           error = xfs_attr3_leaf_flipflags(args);
>>>           if (error)
>>>               return error;
>>> -        /*
>>> -         * Commit the flag value change and start the next trans in
>>> -         * series.
>>> -         */
>>> -        error = xfs_trans_roll_inode(&args->trans, args->dp);
>>> -        if (error)
>>> -            return error;
>>> +
>>> +        args->dac.dela_state = XFS_DAS_FLIP_LFLAG;
>>> +        return -EAGAIN;
>>> +flip_flag:
>>>           /*
>>>            * Dismantle the "old" attribute/value pair by removing
>>> @@ -844,34 +944,17 @@ xfs_attr_leaf_addname(struct xfs_da_args    *args)
>>>           /*
>>>            * If the result is small enough, shrink it all into the 
>>> inode.
>>>            */
>>> -        if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
>>> +        forkoff = xfs_attr_shortform_allfit(bp, dp);
>>> +        if (forkoff)
>>>               error = xfs_attr3_leaf_to_shortform(bp, args, forkoff);
>>> -            /* bp is gone due to xfs_da_shrink_inode */
>>> -            if (error)
>>> -                return error;
>>> -            error = xfs_defer_finish(&args->trans);
>>> -            if (error)
>>> -                return error;
>>> -        }
>>> -        /*
>>> -         * Commit the remove and start the next trans in series.
>>> -         */
>>> -        error = xfs_trans_roll_inode(&args->trans, dp);
>>> +        args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>>       } else if (args->rmtblkno > 0) {
>>>           /*
>>>            * Added a "remote" value, just clear the incomplete flag.
>>>            */
>>>           error = xfs_attr3_leaf_clearflag(args);
>>> -        if (error)
>>> -            return error;
>>> -
>>> -        /*
>>> -         * Commit the flag value change and start the next trans in
>>> -         * series.
>>> -         */
>>> -        error = xfs_trans_roll_inode(&args->trans, args->dp);
>>>       }
>>>       return error;
>>>   }
>>> @@ -1020,16 +1103,22 @@ xfs_attr_node_hasname(
>>>    *
>>>    * "Remote" attribute values confuse the issue and atomic rename 
>>> operations
>>>    * add a whole extra layer of confusion on top of that.
>>> + *
>>> + * This routine is meant to function as a delayed operation, and may 
>>> return
>>> + * -EAGAIN when the transaction needs to be rolled.  Calling 
>>> functions will need
>>> + * to handle this, and recall the function until a successful error 
>>> code is
>>> + *returned.
>>>    */
>>>   STATIC int
>>>   xfs_attr_node_addname(
>>>       struct xfs_da_args    *args)
>>>   {
>>> -    struct xfs_da_state    *state;
>>> +    struct xfs_da_state    *state = NULL;
>>>       struct xfs_da_state_blk    *blk;
>>>       struct xfs_inode    *dp;
>>>       struct xfs_mount    *mp;
>>> -    int            retval, error;
>>> +    int            retval = 0;
>>> +    int            error = 0;
>>>       trace_xfs_attr_node_addname(args);
>>> @@ -1038,7 +1127,19 @@ xfs_attr_node_addname(
>>>        */
>>>       dp = args->dp;
>>>       mp = dp->i_mount;
>>> -restart:
>>> +
>>> +    /* State machine switch */
>>> +    switch (args->dac.dela_state) {
>>> +    case XFS_DAS_FLIP_NFLAG:
>>> +        goto flip_flag;
>>> +    case XFS_DAS_FOUND_NBLK:
>>> +        goto found_nblk;
>>> +    case XFS_DAS_ALLOC_NODE:
>>> +        goto alloc_node;
>>> +    default:
>>> +        break;
>>> +    }
>>> +
>>>       /*
>>>        * Search to see if name already exists, and get back a pointer
>>>        * to where it should go.
>>> @@ -1088,19 +1189,13 @@ xfs_attr_node_addname(
>>>               error = xfs_attr3_leaf_to_node(args);
>>>               if (error)
>>>                   goto out;
>>> -            error = xfs_defer_finish(&args->trans);
>>> -            if (error)
>>> -                goto out;
>>>               /*
>>> -             * Commit the node conversion and start the next
>>> -             * trans in the chain.
>>> +             * Restart routine from the top.  No need to set  the
>>> +             * state
>>
> 
> I think I missed responding to this comment in my last email:
> 
>> *OH*, dela_state == 0 does not mean "do not use delayed attrs", we're
>> switching /all/ users to delayed attrs.  Urk.
> 
> Not quite.  While both inline and delayed operations go through this 
> same dance of EAGAIN ping pong, it's not a delayed operation until we 
> package it into an intent and log it.  And then the intent pops back out 
> during defer finish, and the ping pong starts from 
> xfs_defer_finish_noroll instead of xfs_attr_set_iter.
Sorry, that should say xfs_attr_set_args, not xfs_attr_set_iter.  :-)

> 
> We still need to support in line operations for older versions that do 
> not support the attr log item.  Later when we get into delayed attrs, we 
> use a feature bit to control this, and toggle as needed.
> 
> Hope that helps!
> 
>>
>>>                */
>>> -            error = xfs_trans_roll_inode(&args->trans, dp);
>>> -            if (error)
>>> -                goto out;
>>> -
>>> -            goto restart;
>>> +            args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>> +            return -EAGAIN;
>>>           }
>>>           /*
>>> @@ -1112,9 +1207,7 @@ xfs_attr_node_addname(
>>>           error = xfs_da3_split(state);
>>>           if (error)
>>>               goto out;
>>> -        error = xfs_defer_finish(&args->trans);
>>> -        if (error)
>>> -            goto out;
>>> +        args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>>       } else {
>>>           /*
>>>            * Addition succeeded, update Btree hashvals.
>>> @@ -1129,13 +1222,9 @@ xfs_attr_node_addname(
>>>       xfs_da_state_free(state);
>>>       state = NULL;
>>> -    /*
>>> -     * Commit the leaf addition or btree split and start the next
>>> -     * trans in the chain.
>>> -     */
>>> -    error = xfs_trans_roll_inode(&args->trans, dp);
>>> -    if (error)
>>> -        goto out;
>>> +    args->dac.dela_state = XFS_DAS_FOUND_NBLK;
>>> +    return -EAGAIN;
>>> +found_nblk:
>>>       /*
>>>        * If there was an out-of-line value, allocate the blocks we
>>> @@ -1144,7 +1233,27 @@ xfs_attr_node_addname(
>>>        * maximum size of a transaction and/or hit a deadlock.
>>>        */
>>>       if (args->rmtblkno > 0) {
>>> -        error = xfs_attr_rmtval_set(args);
>>> +        /* Open coded xfs_attr_rmtval_set without trans handling */
>>> +        error = xfs_attr_rmtval_set_init(args);
>>> +        if (error)
>>> +            return error;
>>> +
>>> +        /*
>>> +         * Roll through the "value", allocating blocks on disk as
>>> +         * required.
>>> +         */
>>> +alloc_node:
>>> +        while (args->dac.blkcnt > 0) {
>>> +            error = xfs_attr_rmtval_set_blk(args);
>>> +            if (error)
>>> +                return error;
>>> +
>>> +            args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>> +            args->dac.dela_state = XFS_DAS_ALLOC_NODE;
>>> +            return -EAGAIN;
>>> +        }
>>> +
>>> +        error = xfs_attr_rmtval_set_value(args);
>>>           if (error)
>>>               return error;
>>>       }
>>> @@ -1167,10 +1276,9 @@ xfs_attr_node_addname(
>>>            * Commit the flag value change and start the next trans in
>>>            * series
>>>            */
>>> -        error = xfs_trans_roll_inode(&args->trans, args->dp);
>>> -        if (error)
>>> -            goto out;
>>> -
>>> +        args->dac.dela_state = XFS_DAS_FLIP_NFLAG;
>>> +        return -EAGAIN;
>>> +flip_flag:
>>>           /*
>>>            * Dismantle the "old" attribute/value pair by removing
>>>            * a "remote" value (if it exists).
>>> @@ -1199,7 +1307,6 @@ xfs_attr_node_addname(
>>>           error = xfs_da3_node_lookup_int(state, &retval);
>>>           if (error)
>>>               goto out;
>>> -
>>>           /*
>>>            * Remove the name and update the hashvals in the tree.
>>>            */
>>> @@ -1207,7 +1314,6 @@ xfs_attr_node_addname(
>>>           ASSERT(blk->magic == XFS_ATTR_LEAF_MAGIC);
>>>           error = xfs_attr3_leaf_remove(blk->bp, args);
>>>           xfs_da3_fixhashpath(state, &state->path);
>>> -
>>>           /*
>>>            * Check to see if the tree needs to be collapsed.
>>>            */
>>> @@ -1215,18 +1321,9 @@ xfs_attr_node_addname(
>>>               error = xfs_da3_join(state);
>>>               if (error)
>>>                   goto out;
>>> -            error = xfs_defer_finish(&args->trans);
>>> -            if (error)
>>> -                goto out;
>>> -        }
>>> -
>>> -        /*
>>> -         * Commit and start the next trans in the chain.
>>> -         */
>>> -        error = xfs_trans_roll_inode(&args->trans, dp);
>>> -        if (error)
>>> -            goto out;
>>> +            args->dac.flags |= XFS_DAC_FINISH_TRANS;
>>> +        }
>>>       } else if (args->rmtblkno > 0) {
>>>           /*
>>>            * Added a "remote" value, just clear the incomplete flag.
>>> @@ -1234,14 +1331,6 @@ xfs_attr_node_addname(
>>>           error = xfs_attr3_leaf_clearflag(args);
>>>           if (error)
>>>               goto out;
>>> -
>>> -         /*
>>> -          * Commit the flag value change and start the next trans in
>>> -          * series.
>>> -          */
>>> -        error = xfs_trans_roll_inode(&args->trans, args->dp);
>>> -        if (error)
>>> -            goto out;
>>>       }
>>>       retval = error = 0;
>>> diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
>>> index ea873a5..f450d8c 100644
>>> --- a/fs/xfs/libxfs/xfs_attr.h
>>> +++ b/fs/xfs/libxfs/xfs_attr.h
>>> @@ -152,6 +152,7 @@ int xfs_attr_get(struct xfs_inode *ip, struct 
>>> xfs_name *name,
>>>   int xfs_attr_set(struct xfs_inode *dp, struct xfs_name *name,
>>>            unsigned char *value, int valuelen, int flags);
>>>   int xfs_attr_set_args(struct xfs_da_args *args);
>>> +int xfs_attr_set_iter(struct xfs_da_args *args, struct xfs_buf 
>>> **leaf_bp);
>>>   int xfs_attr_remove(struct xfs_inode *dp, struct xfs_name *name, 
>>> int flags);
>>>   int xfs_has_attr(struct xfs_da_args *args);
>>>   int xfs_attr_remove_args(struct xfs_da_args *args);
>>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.c 
>>> b/fs/xfs/libxfs/xfs_attr_remote.c
>>> index f2ee0b8..ad4f14a 100644
>>> --- a/fs/xfs/libxfs/xfs_attr_remote.c
>>> +++ b/fs/xfs/libxfs/xfs_attr_remote.c
>>> @@ -428,7 +428,7 @@ xfs_attr_rmtval_get(
>>>    * Find a "hole" in the attribute address space large enough for us 
>>> to drop the
>>>    * new attribute's value into
>>>    */
>>> -STATIC int
>>> +int
>>>   xfs_attr_rmt_find_hole(
>>>       struct xfs_da_args    *args)
>>>   {
>>> @@ -455,7 +455,7 @@ xfs_attr_rmt_find_hole(
>>>       return 0;
>>>   }
>>> -STATIC int
>>> +int
>>>   xfs_attr_rmtval_set_value(
>>>       struct xfs_da_args    *args)
>>>   {
>>> @@ -588,6 +588,69 @@ xfs_attr_rmtval_set(
>>>   }
>>>   /*
>>> + * Find a hole for the attr and store it in the delayed attr 
>>> context.  This
>>> + * initializes the context to roll through allocating an attr extent 
>>> for a
>>> + * delayed attr operation
>>> + */
>>> +int
>>> +xfs_attr_rmtval_set_init(
>>> +    struct xfs_da_args    *args)
>>> +{
>>> +    struct xfs_bmbt_irec    *map = &args->dac.map;
>>> +    int error;
>>> +
>>> +    args->dac.lblkno = 0;
>>> +    args->dac.lfileoff = 0;
>>> +    args->dac.blkcnt = 0;
>>> +    args->rmtblkcnt = 0;
>>> +    args->rmtblkno = 0;
>>> +    memset(map, 0, sizeof(struct xfs_bmbt_irec));
>>> +
>>> +    error = xfs_attr_rmt_find_hole(args);
>>> +    if (error)
>>> +        return error;
>>> +
>>> +    args->dac.blkcnt = args->rmtblkcnt;
>>> +    args->dac.lblkno = args->rmtblkno;
>>> +
>>> +    return error;
>>> +}
>>> +
>>> +/*
>>> + * Write one block of the value associated with an attribute into the
>>> + * out-of-line buffer that we have defined for it. This is similar 
>>> to a subset
>>> + * of xfs_attr_rmtval_set, but records the current block to the 
>>> delayed attr
>>> + * context, and leaves transaction handling to the caller.
>>> + */
>>> +int
>>> +xfs_attr_rmtval_set_blk(
>>> +    struct xfs_da_args    *args)
>>> +{
>>> +    struct xfs_inode    *dp = args->dp;
>>> +    struct xfs_bmbt_irec    *map = &args->dac.map;
>>> +    int nmap;
>>> +    int error;
>>> +
>>> +    nmap = 1;
>>> +    error = xfs_bmapi_write(args->trans, dp,
>>> +          (xfs_fileoff_t)args->dac.lblkno,
>>> +          args->dac.blkcnt, XFS_BMAPI_ATTRFORK,
>>> +          args->total, map, &nmap);
>>> +    if (error)
>>> +        return error;
>>> +
>>> +    ASSERT(nmap == 1);
>>> +    ASSERT((map->br_startblock != DELAYSTARTBLOCK) &&
>>> +           (map->br_startblock != HOLESTARTBLOCK));
>>> +
>>> +    /* roll attribute extent map forwards */
>>> +    args->dac.lblkno += map->br_blockcount;
>>> +    args->dac.blkcnt -= map->br_blockcount;
>>> +
>>> +    return 0;
>>> +}
>>> +
>>> +/*
>>>    * Remove the value associated with an attribute by deleting the
>>>    * out-of-line buffer that it is stored on.
>>>    */
>>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.h 
>>> b/fs/xfs/libxfs/xfs_attr_remote.h
>>> index 7ab3770..ab03519 100644
>>> --- a/fs/xfs/libxfs/xfs_attr_remote.h
>>> +++ b/fs/xfs/libxfs/xfs_attr_remote.h
>>> @@ -13,4 +13,8 @@ int xfs_attr_rmtval_set(struct xfs_da_args *args);
>>>   int xfs_attr_rmtval_remove(struct xfs_da_args *args);
>>>   int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
>>>   int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
>>> +int xfs_attr_rmt_find_hole(struct xfs_da_args *args);
>>> +int xfs_attr_rmtval_set_value(struct xfs_da_args *args);
>>> +int xfs_attr_rmtval_set_blk(struct xfs_da_args *args);
>>> +int xfs_attr_rmtval_set_init(struct xfs_da_args *args);
>>>   #endif /* __XFS_ATTR_REMOTE_H__ */
>>> diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
>>> index 7fc87da..9943062 100644
>>> --- a/fs/xfs/libxfs/xfs_da_btree.h
>>> +++ b/fs/xfs/libxfs/xfs_da_btree.h
>>> @@ -55,6 +55,14 @@ enum xfs_dacmp {
>>>   enum xfs_delattr_state {
>>>       XFS_DAS_RM_SHRINK    = 1, /* We are shrinking the tree */
>>>       XFS_DAS_RM_NODE_BLKS    = 2, /* We are removing node blocks */
>>> +    XFS_DAS_SF_TO_LEAF    = 3, /* Converted short form to leaf */
>>> +    XFS_DAS_FOUND_LBLK    = 4, /* We found leaf blk for attr */
>>> +    XFS_DAS_LEAF_TO_NODE    = 5, /* Converted leaf to node */
>>> +    XFS_DAS_FOUND_NBLK    = 6, /* We found node blk for attr */
>>> +    XFS_DAS_ALLOC_LEAF    = 7, /* We are allocating leaf blocks */
>>> +    XFS_DAS_FLIP_LFLAG    = 8, /* Flipped leaf INCOMPLETE attr flag */
>>> +    XFS_DAS_ALLOC_NODE    = 9, /* We are allocating node blocks */
>>> +    XFS_DAS_FLIP_NFLAG    = 10,/* Flipped node INCOMPLETE attr flag */
>>
>> We've definitely reached the point where a state diagram would be
>> helpful.  Can you go from any of the RM_ states to the ones that you've
>> just added?
>>
>> The new state machine code in last two patches would be a lot easier to
>> review if I could look down from above instead of up from the XFS_DAS
>> values and goto labels.  I /think/ it looks sane, but I'm only 20%
>> confident of that statement.
>>
>> --D
>>
>>>   };
>>>   /*
>>> @@ -66,8 +74,13 @@ enum xfs_delattr_state {
>>>    * Context used for keeping track of delayed attribute operations
>>>    */
>>>   struct xfs_delattr_context {
>>> +    struct xfs_bmbt_irec    map;
>>> +    struct xfs_buf        *leaf_bp;
>>> +    xfs_fileoff_t        lfileoff;
>>>       struct xfs_da_state    *da_state;
>>>       struct xfs_da_state_blk *blk;
>>> +    xfs_dablk_t        lblkno;
>>> +    int            blkcnt;
>>>       int            flags;
>>>       enum xfs_delattr_state    dela_state;
>>>   };
>>> -- 
>>> 2.7.4
>>>

Re: [PATCH v6 11/16] xfs: Check for -ENOATTR or -EEXIST

[Allison Collins]

On 1/21/20 9:29 PM, Allison Collins wrote:
> 
> 
> On 1/21/20 4:15 PM, Darrick J. Wong wrote:
>> On Sat, Jan 18, 2020 at 03:50:30PM -0700, Allison Collins wrote:
>>> Delayed operations cannot return error codes.  So we must check for
>>> these conditions first before starting set or remove operations
>>
>> Answering my own question from earlier -- I see here you actually /are/
>> checking the attr existence w.r.t. ATTR_{CREATE,REPLACE} right after we
>> allocate a transaction and ILOCK the inode, so
>>
>> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
> Alrighty, thank you!
> 
>>
>> Though I am wondering if you could discard the predicates from the
>> second patch in favor of doing a normal lookup of the attr with a zero
>> valuelen to determine if there's already an attribute?
> I think I likely answered this in the response to that patch.  Because 
> it's used as part of the remove procedures, we still need it.  We could 
> make a simpler version just for this application I suppose, but it seems 
> like it'd just be extra code since we still need the former.
> 
> Thank you for the reviews!
> Allison
> 
>>
>> --D
>>
>>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>>> ---
>>>   fs/xfs/libxfs/xfs_attr.c | 12 ++++++++++++
>>>   1 file changed, 12 insertions(+)
>>>
>>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>>> index a2673fe..e9d22c1 100644
>>> --- a/fs/xfs/libxfs/xfs_attr.c
>>> +++ b/fs/xfs/libxfs/xfs_attr.c
>>> @@ -457,6 +457,14 @@ xfs_attr_set(
>>>           goto out_trans_cancel;
>>>       xfs_trans_ijoin(args.trans, dp, 0);
>>> +
>>> +    error = xfs_has_attr(&args);
>>> +    if (error == -EEXIST && (name->type & ATTR_CREATE))
>>> +        goto out_trans_cancel;
>>> +
>>> +    if (error == -ENOATTR && (name->type & ATTR_REPLACE))
>>> +        goto out_trans_cancel;
>>> +
>>>       error = xfs_attr_set_args(&args);
So I was thinking of adding this one to a smaller 3 patch series I 
mentioned earlier.  I was also thinking of adding in some asserts here:

ASSERT(error != -EEXIST)
ASSERT(error != -ENOATTR)

Just to make sure the changes are enforcing the behavioral changes that 
we want.  I thought this might be a good stabilizer to the rest of the 
delayed attr series.  Because chasing this bug back up through the log 
replay is a much bigger PITA than catching it here.  Thoughts?

>>>       if (error)
>>>           goto out_trans_cancel;
>>> @@ -545,6 +553,10 @@ xfs_attr_remove(
>>>        */
>>>       xfs_trans_ijoin(args.trans, dp, 0);
>>> +    error = xfs_has_attr(&args);
>>> +    if (error != -EEXIST)
>>> +        goto out;
>>> +
Here too:
ASSERT(error != -EEXIST)

Let me know what folks think.  Thanks!

Allison

>>>       error = xfs_attr_remove_args(&args);
>>>       if (error)
>>>           goto out;
>>> -- 
>>> 2.7.4
>>>

Re: [PATCH v6 11/16] xfs: Check for -ENOATTR or -EEXIST

[Darrick J. Wong]

On Sat, Jan 25, 2020 at 09:41:47AM -0700, Allison Collins wrote:
> On 1/21/20 9:29 PM, Allison Collins wrote:
> > 
> > 
> > On 1/21/20 4:15 PM, Darrick J. Wong wrote:
> > > On Sat, Jan 18, 2020 at 03:50:30PM -0700, Allison Collins wrote:
> > > > Delayed operations cannot return error codes.  So we must check for
> > > > these conditions first before starting set or remove operations
> > > 
> > > Answering my own question from earlier -- I see here you actually /are/
> > > checking the attr existence w.r.t. ATTR_{CREATE,REPLACE} right after we
> > > allocate a transaction and ILOCK the inode, so
> > > 
> > > Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
> > Alrighty, thank you!
> > 
> > > 
> > > Though I am wondering if you could discard the predicates from the
> > > second patch in favor of doing a normal lookup of the attr with a zero
> > > valuelen to determine if there's already an attribute?
> > I think I likely answered this in the response to that patch.  Because
> > it's used as part of the remove procedures, we still need it.  We could
> > make a simpler version just for this application I suppose, but it seems
> > like it'd just be extra code since we still need the former.
> > 
> > Thank you for the reviews!
> > Allison
> > 
> > > 
> > > --D
> > > 
> > > > Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> > > > ---
> > > >   fs/xfs/libxfs/xfs_attr.c | 12 ++++++++++++
> > > >   1 file changed, 12 insertions(+)
> > > > 
> > > > diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> > > > index a2673fe..e9d22c1 100644
> > > > --- a/fs/xfs/libxfs/xfs_attr.c
> > > > +++ b/fs/xfs/libxfs/xfs_attr.c
> > > > @@ -457,6 +457,14 @@ xfs_attr_set(
> > > >           goto out_trans_cancel;
> > > >       xfs_trans_ijoin(args.trans, dp, 0);
> > > > +
> > > > +    error = xfs_has_attr(&args);
> > > > +    if (error == -EEXIST && (name->type & ATTR_CREATE))
> > > > +        goto out_trans_cancel;
> > > > +
> > > > +    if (error == -ENOATTR && (name->type & ATTR_REPLACE))
> > > > +        goto out_trans_cancel;
> > > > +
> > > >       error = xfs_attr_set_args(&args);
> So I was thinking of adding this one to a smaller 3 patch series I mentioned
> earlier.  I was also thinking of adding in some asserts here:
> 
> ASSERT(error != -EEXIST)
> ASSERT(error != -ENOATTR)
> 
> Just to make sure the changes are enforcing the behavioral changes that we
> want.  I thought this might be a good stabilizer to the rest of the delayed
> attr series.  Because chasing this bug back up through the log replay is a
> much bigger PITA than catching it here.  Thoughts?

Er, are the asserts to check that xfs_attr_set_args never returns
EEXIST/ENOATTR?  I'm not sure why you'd have to chase this through log
replay?

/me is in this funny place where he thinks that in general adding
asserts (or WARN_ON) to check assumptions is a good idea, but not sure
what's going on here.

--D

> > > >       if (error)
> > > >           goto out_trans_cancel;
> > > > @@ -545,6 +553,10 @@ xfs_attr_remove(
> > > >        */
> > > >       xfs_trans_ijoin(args.trans, dp, 0);
> > > > +    error = xfs_has_attr(&args);
> > > > +    if (error != -EEXIST)
> > > > +        goto out;
> > > > +
> Here too:
> ASSERT(error != -EEXIST)
> 
> Let me know what folks think.  Thanks!
> 
> Allison
> 
> > > >       error = xfs_attr_remove_args(&args);
> > > >       if (error)
> > > >           goto out;
> > > > -- 
> > > > 2.7.4
> > > > 

Re: [PATCH v6 11/16] xfs: Check for -ENOATTR or -EEXIST

[Allison Collins]

On 1/26/20 3:28 PM, Darrick J. Wong wrote:
> On Sat, Jan 25, 2020 at 09:41:47AM -0700, Allison Collins wrote:
>> On 1/21/20 9:29 PM, Allison Collins wrote:
>>>
>>>
>>> On 1/21/20 4:15 PM, Darrick J. Wong wrote:
>>>> On Sat, Jan 18, 2020 at 03:50:30PM -0700, Allison Collins wrote:
>>>>> Delayed operations cannot return error codes.  So we must check for
>>>>> these conditions first before starting set or remove operations
>>>>
>>>> Answering my own question from earlier -- I see here you actually /are/
>>>> checking the attr existence w.r.t. ATTR_{CREATE,REPLACE} right after we
>>>> allocate a transaction and ILOCK the inode, so
>>>>
>>>> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
>>> Alrighty, thank you!
>>>
>>>>
>>>> Though I am wondering if you could discard the predicates from the
>>>> second patch in favor of doing a normal lookup of the attr with a zero
>>>> valuelen to determine if there's already an attribute?
>>> I think I likely answered this in the response to that patch.  Because
>>> it's used as part of the remove procedures, we still need it.  We could
>>> make a simpler version just for this application I suppose, but it seems
>>> like it'd just be extra code since we still need the former.
>>>
>>> Thank you for the reviews!
>>> Allison
>>>
>>>>
>>>> --D
>>>>
>>>>> Signed-off-by: Allison Collins <allison.henderson@oracle.com>
>>>>> ---
>>>>>    fs/xfs/libxfs/xfs_attr.c | 12 ++++++++++++
>>>>>    1 file changed, 12 insertions(+)
>>>>>
>>>>> diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
>>>>> index a2673fe..e9d22c1 100644
>>>>> --- a/fs/xfs/libxfs/xfs_attr.c
>>>>> +++ b/fs/xfs/libxfs/xfs_attr.c
>>>>> @@ -457,6 +457,14 @@ xfs_attr_set(
>>>>>            goto out_trans_cancel;
>>>>>        xfs_trans_ijoin(args.trans, dp, 0);
>>>>> +
>>>>> +    error = xfs_has_attr(&args);
>>>>> +    if (error == -EEXIST && (name->type & ATTR_CREATE))
>>>>> +        goto out_trans_cancel;
>>>>> +
>>>>> +    if (error == -ENOATTR && (name->type & ATTR_REPLACE))
>>>>> +        goto out_trans_cancel;
>>>>> +
>>>>>        error = xfs_attr_set_args(&args);
>> So I was thinking of adding this one to a smaller 3 patch series I mentioned
>> earlier.  I was also thinking of adding in some asserts here:
>>
>> ASSERT(error != -EEXIST)
>> ASSERT(error != -ENOATTR)
>>
>> Just to make sure the changes are enforcing the behavioral changes that we
>> want.  I thought this might be a good stabilizer to the rest of the delayed
>> attr series.  Because chasing this bug back up through the log replay is a
>> much bigger PITA than catching it here.  Thoughts?
> 
> Er, are the asserts to check that xfs_attr_set_args never returns
> EEXIST/ENOATTR?  I'm not sure why you'd have to chase this through log
> replay?

Yes, the idea is that EEXIST and ENOATTR are supposed to be found and 
returned by the xfs_has_attr routine above. If they happen at this 
point, it would actually be more of an internal error.  For example: if 
we're renaming an attr, and xfs_has_attr finds that it exists, but then 
xfs_attr_set_args comes back with ENOATTR, clearly something unexpected 
happened.

The motivation for this is just that if it does happen, it's easier to 
work out this out now rather than later when we bring in the rest of the 
delayed attribute code.  Because in that case, the error wont happen 
here, it will happen later as part of a finish_item (or a log replay).

It's not a requirement I suppose, just more of a pro-active check really.

> 
> /me is in this funny place where he thinks that in general adding
> asserts (or WARN_ON) to check assumptions is a good idea, but not sure
> what's going on here.
Did that answer your question then?

> 
> --D
> 
>>>>>        if (error)
>>>>>            goto out_trans_cancel;
>>>>> @@ -545,6 +553,10 @@ xfs_attr_remove(
>>>>>         */
>>>>>        xfs_trans_ijoin(args.trans, dp, 0);
>>>>> +    error = xfs_has_attr(&args);
>>>>> +    if (error != -EEXIST)
>>>>> +        goto out;
>>>>> +
>> Here too:
>> ASSERT(error != -EEXIST)
>>
>> Let me know what folks think.  Thanks!
>>
>> Allison
>>
>>>>>        error = xfs_attr_remove_args(&args);
>>>>>        if (error)
>>>>>            goto out;
>>>>> -- 
>>>>> 2.7.4
>>>>>

Re: [PATCH v6 11/16] xfs: Check for -ENOATTR or -EEXIST

[Darrick J. Wong]

On Sun, Jan 26, 2020 at 05:20:16PM -0700, Allison Collins wrote:
> 
> 
> On 1/26/20 3:28 PM, Darrick J. Wong wrote:
> > On Sat, Jan 25, 2020 at 09:41:47AM -0700, Allison Collins wrote:
> > > On 1/21/20 9:29 PM, Allison Collins wrote:
> > > > 
> > > > 
> > > > On 1/21/20 4:15 PM, Darrick J. Wong wrote:
> > > > > On Sat, Jan 18, 2020 at 03:50:30PM -0700, Allison Collins wrote:
> > > > > > Delayed operations cannot return error codes.  So we must check for
> > > > > > these conditions first before starting set or remove operations
> > > > > 
> > > > > Answering my own question from earlier -- I see here you actually /are/
> > > > > checking the attr existence w.r.t. ATTR_{CREATE,REPLACE} right after we
> > > > > allocate a transaction and ILOCK the inode, so
> > > > > 
> > > > > Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
> > > > Alrighty, thank you!
> > > > 
> > > > > 
> > > > > Though I am wondering if you could discard the predicates from the
> > > > > second patch in favor of doing a normal lookup of the attr with a zero
> > > > > valuelen to determine if there's already an attribute?
> > > > I think I likely answered this in the response to that patch.  Because
> > > > it's used as part of the remove procedures, we still need it.  We could
> > > > make a simpler version just for this application I suppose, but it seems
> > > > like it'd just be extra code since we still need the former.
> > > > 
> > > > Thank you for the reviews!
> > > > Allison
> > > > 
> > > > > 
> > > > > --D
> > > > > 
> > > > > > Signed-off-by: Allison Collins <allison.henderson@oracle.com>
> > > > > > ---
> > > > > >    fs/xfs/libxfs/xfs_attr.c | 12 ++++++++++++
> > > > > >    1 file changed, 12 insertions(+)
> > > > > > 
> > > > > > diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> > > > > > index a2673fe..e9d22c1 100644
> > > > > > --- a/fs/xfs/libxfs/xfs_attr.c
> > > > > > +++ b/fs/xfs/libxfs/xfs_attr.c
> > > > > > @@ -457,6 +457,14 @@ xfs_attr_set(
> > > > > >            goto out_trans_cancel;
> > > > > >        xfs_trans_ijoin(args.trans, dp, 0);
> > > > > > +
> > > > > > +    error = xfs_has_attr(&args);
> > > > > > +    if (error == -EEXIST && (name->type & ATTR_CREATE))
> > > > > > +        goto out_trans_cancel;
> > > > > > +
> > > > > > +    if (error == -ENOATTR && (name->type & ATTR_REPLACE))
> > > > > > +        goto out_trans_cancel;
> > > > > > +
> > > > > >        error = xfs_attr_set_args(&args);
> > > So I was thinking of adding this one to a smaller 3 patch series I mentioned
> > > earlier.  I was also thinking of adding in some asserts here:
> > > 
> > > ASSERT(error != -EEXIST)
> > > ASSERT(error != -ENOATTR)
> > > 
> > > Just to make sure the changes are enforcing the behavioral changes that we
> > > want.  I thought this might be a good stabilizer to the rest of the delayed
> > > attr series.  Because chasing this bug back up through the log replay is a
> > > much bigger PITA than catching it here.  Thoughts?
> > 
> > Er, are the asserts to check that xfs_attr_set_args never returns
> > EEXIST/ENOATTR?  I'm not sure why you'd have to chase this through log
> > replay?
> 
> Yes, the idea is that EEXIST and ENOATTR are supposed to be found and
> returned by the xfs_has_attr routine above. If they happen at this point, it
> would actually be more of an internal error.  For example: if we're renaming
> an attr, and xfs_has_attr finds that it exists, but then xfs_attr_set_args
> comes back with ENOATTR, clearly something unexpected happened.
> 
> The motivation for this is just that if it does happen, it's easier to work
> out this out now rather than later when we bring in the rest of the delayed
> attribute code.  Because in that case, the error wont happen here, it will
> happen later as part of a finish_item (or a log replay).
> 
> It's not a requirement I suppose, just more of a pro-active check really.

Ok, it's purely a defensive check so that you'll notice problems early
before the fs goes bonkers, not a problem that mysteriously appears but
only after some random unexpected shutdown.

> > 
> > /me is in this funny place where he thinks that in general adding
> > asserts (or WARN_ON) to check assumptions is a good idea, but not sure
> > what's going on here.
> Did that answer your question then?

Yep.

--D

> > 
> > --D
> > 
> > > > > >        if (error)
> > > > > >            goto out_trans_cancel;
> > > > > > @@ -545,6 +553,10 @@ xfs_attr_remove(
> > > > > >         */
> > > > > >        xfs_trans_ijoin(args.trans, dp, 0);
> > > > > > +    error = xfs_has_attr(&args);
> > > > > > +    if (error != -EEXIST)
> > > > > > +        goto out;
> > > > > > +
> > > Here too:
> > > ASSERT(error != -EEXIST)
> > > 
> > > Let me know what folks think.  Thanks!
> > > 
> > > Allison
> > > 
> > > > > >        error = xfs_attr_remove_args(&args);
> > > > > >        if (error)
> > > > > >            goto out;
> > > > > > -- 
> > > > > > 2.7.4
> > > > > > 

Re: [PATCH v6 16/16] xfs: Add delay ready attr set routines

[Allison Collins]

On 1/22/20 3:30 AM, Allison Collins wrote:
> 
> 
> On 1/21/20 5:12 PM, Darrick J. Wong wrote:
>> On Sat, Jan 18, 2020 at 03:50:35PM -0700, Allison Collins wrote:

<snip>

>>> diff --git a/fs/xfs/libxfs/xfs_attr_remote.h 
>>> b/fs/xfs/libxfs/xfs_attr_remote.h
>>> index 7ab3770..ab03519 100644
>>> --- a/fs/xfs/libxfs/xfs_attr_remote.h
>>> +++ b/fs/xfs/libxfs/xfs_attr_remote.h
>>> @@ -13,4 +13,8 @@ int xfs_attr_rmtval_set(struct xfs_da_args *args);
>>>   int xfs_attr_rmtval_remove(struct xfs_da_args *args);
>>>   int xfs_attr_rmtval_invalidate(struct xfs_da_args *args);
>>>   int xfs_attr_rmtval_unmap(struct xfs_da_args *args);
>>> +int xfs_attr_rmt_find_hole(struct xfs_da_args *args);
>>> +int xfs_attr_rmtval_set_value(struct xfs_da_args *args);
>>> +int xfs_attr_rmtval_set_blk(struct xfs_da_args *args);
>>> +int xfs_attr_rmtval_set_init(struct xfs_da_args *args);
>>>   #endif /* __XFS_ATTR_REMOTE_H__ */
>>> diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
>>> index 7fc87da..9943062 100644
>>> --- a/fs/xfs/libxfs/xfs_da_btree.h
>>> +++ b/fs/xfs/libxfs/xfs_da_btree.h
>>> @@ -55,6 +55,14 @@ enum xfs_dacmp {
>>>   enum xfs_delattr_state {
>>>       XFS_DAS_RM_SHRINK    = 1, /* We are shrinking the tree */
>>>       XFS_DAS_RM_NODE_BLKS    = 2, /* We are removing node blocks */
>>> +    XFS_DAS_SF_TO_LEAF    = 3, /* Converted short form to leaf */
>>> +    XFS_DAS_FOUND_LBLK    = 4, /* We found leaf blk for attr */
>>> +    XFS_DAS_LEAF_TO_NODE    = 5, /* Converted leaf to node */
>>> +    XFS_DAS_FOUND_NBLK    = 6, /* We found node blk for attr */
>>> +    XFS_DAS_ALLOC_LEAF    = 7, /* We are allocating leaf blocks */
>>> +    XFS_DAS_FLIP_LFLAG    = 8, /* Flipped leaf INCOMPLETE attr flag */
>>> +    XFS_DAS_ALLOC_NODE    = 9, /* We are allocating node blocks */
>>> +    XFS_DAS_FLIP_NFLAG    = 10,/* Flipped node INCOMPLETE attr flag */
>>
>> We've definitely reached the point where a state diagram would be
>> helpful.  Can you go from any of the RM_ states to the ones that you've
>> just added?
> No.  And if they did, they would have to show up in the calling 
> functions state switches.  Because the calling function has to manage 
> jumping straight back to the subroutine when ever the state belongs to 
> the subroutine.  Kind of like how you see XFS_DAS_ALLOC_LEAF and 
> XFS_DAS_FLIP_LFLAG in the state switch for xfs_attr_set_iter, even 
> though those states only apply to xfs_attr_leaf_addname.
> 
>>
>> The new state machine code in last two patches would be a lot easier to
>> review if I could look down from above instead of up from the XFS_DAS
>> values and goto labels.  I /think/ it looks sane, but I'm only 20%
>> confident of that statement.
>>
> Sure, I'll see if I can put together a diagram to help folks out a bit.
> 
> Thanks for the reviews!
> Allison
> 
>> --D
>>

Ok, so I've put together some quick high level diagrams, and I wanted to 
run it by you to see what you thought.

These diagrams illustrate the state machine logic as it appears in the 
current set. The XFS_DAS_* states indicate places where the function 
would return -EAGAIN, and then immediately resume from after being 
recalled by the calling function.  States marked as a "subroutine state" 
indicate that they belong to a subroutine, and so the calling function 
needs to pass them back to that subroutine to allow it to finish where 
it left off.  But they otherwise do not have a role in the calling 
function other than just passing through.

Hope this helps!  Let me know if you have any questions or if there's 
anything that would help make things more clear.  Thanks!

Allison

/*
  * State machine diagram for attr remove operations
  *
  * xfs_attr_remove_iter()
  *         XFS_DAS_RM_SHRINK     ─┐
  *         (subroutine state)     │
  *                                │
  *         XFS_DAS_RMTVAL_REMOVE ─┤
  *         (subroutine state)     │
  *                                └─>xfs_attr_node_removename()
  *                                                 │
  *                                                 v
  *                                         need to remove
  *                                   ┌─n──  rmt blocks?
  *                                   │             │
  *                                   │             y
  *                                   │             │
  *                                   │             v
  *                                   │  ┌─>XFS_DAS_RMTVAL_REMOVE
  *                                   │  │          │
  *                                   │  │          v
  *                                   │  └──y── more blks
  *                                   │         to remove?
  *                                   │             │
  *                                   │             n
  *                                   │             │
  *                                   │             v
  *                                   │         need to
  *                                   └─────> shrink tree? ─n─┐
  *                                                 │         │
  *                                                 y         │
  *                                                 │         │
  *                                                 v         │
  *                                         XFS_DAS_RM_SHRINK │
  *                                                 │         │
  *                                                 v         │
  *                                                done <─────┘
  */

/*
  * State machine diagram for attr set operations
  *
  * xfs_attr_set_iter()
  *                 │
  *                 v
  *           need to upgrade
  *          from sf to leaf? ──n─┐
  *                 │             │
  *                 y             │
  *                 │             │
  *                 V             │
  *          XFS_DAS_ADD_LEAF     │
  *                 │             │
  *                 v             │
  *  ┌──────n── fork has   <──────┘
  *  │         only 1 blk?
  *  │              │
  *  │              y
  *  │              │
  *  │              v
  *  │     xfs_attr_leaf_try_add()
  *  │              │
  *  │              v
  *  │          had enough
  *  ├──────n──   space?
  *  │              │
  *  │              y
  *  │              │
  *  │              v
  *  │      XFS_DAS_FOUND_LBLK  ──┐
  *  │                            │
  *  │      XFS_DAS_FLIP_LFLAG  ──┤
  *  │      (subroutine state)    │
  *  │                            │
  *  │      XFS_DAS_ALLOC_LEAF  ──┤
  *  │      (subroutine state)    │
  *  │                            └─>xfs_attr_leaf_addname()
  *  │                                              │
  *  │                                              v
  *  │                                ┌─────n──  need to
  *  │                                │        alloc blks?
  *  │                                │             │
  *  │                                │             y
  *  │                                │             │
  *  │                                │             v
  *  │                                │  ┌─>XFS_DAS_ALLOC_LEAF
  *  │                                │  │          │
  *  │                                │  │          v
  *  │                                │  └──y── need to alloc
  *  │                                │         more blocks?
  *  │                                │             │
  *  │                                │             n
  *  │                                │             │
  *  │                                │             v
  *  │                                │          was this
  *  │                                └────────> a rename? ──n─┐
  *  │                                              │          │
  *  │                                              y          │
  *  │                                              │          │
  *  │                                              v          │
  *  │                                      XFS_DAS_FLIP_LFLAG │
  *  │                                              │          │
  *  │                                              v          │
  *  │                                             done <──────┘
  *  └────> XFS_DAS_LEAF_TO_NODE ─┐
  *                               │
  *         XFS_DAS_FOUND_NBLK  ──┤
  *         (subroutine state)    │
  *                               │
  *         XFS_DAS_ALLOC_NODE  ──┤
  *         (subroutine state)    │
  *                               │
  *         XFS_DAS_FLIP_NFLAG  ──┤
  *         (subroutine state)    │
  *                               │
  *                               └─>xfs_attr_node_addname()
  *                                                 │
  *                                                 v
  *                                         find space to store
  *                                        attr. Split if needed
  *                                                 │
  *                                                 v
  *                                         XFS_DAS_FOUND_NBLK
  *                                                 │
  *                                                 v
  *                                   ┌─────n──  need to
  *                                   │        alloc blks?
  *                                   │             │
  *                                   │             y
  *                                   │             │
  *                                   │             v
  *                                   │  ┌─>XFS_DAS_ALLOC_NODE
  *                                   │  │          │
  *                                   │  │          v
  *                                   │  └──y── need to alloc
  *                                   │         more blocks?
  *                                   │             │
  *                                   │             n
  *                                   │             │
  *                                   │             v
  *                                   │          was this
  *                                   └────────> a rename? ──n─┐
  *                                                 │          │
  *                                                 y          │
  *                                                 │          │
  *                                                 v          │
  *                                         XFS_DAS_FLIP_NFLAG │
  *                                                 │          │
  *                                                 v          │
  *                                                done <──────┘
  */