[PATCH 0/2] Cleanup for some hardcoded constants

[PATCH 0/2] Cleanup for some hardcoded constants

[ednadolski]

From: Edmund Nadolski <enadolski@suse.com>

This series replaces several hard-coded values with descriptive
symbols.

Edmund Nadolski (2):
  btrfs: provide enumeration for __merge_refs mode argument
  btrfs: replace hardcoded value with SEQ_NONE macro

 fs/btrfs/backref.c | 31 +++++++++++++++++--------------
 fs/btrfs/backref.h |  2 ++
 fs/btrfs/qgroup.c  |  4 ++--
 3 files changed, 21 insertions(+), 16 deletions(-)

-- 
2.10.2

[PATCH 1/2] btrfs: provide enumeration for __merge_refs mode argument

[ednadolski]

From: Edmund Nadolski <enadolski@suse.com>

Replace hardcoded numeric values for __merge_refs 'mode' argument
with descriptive constants.

Signed-off-by: Edmund Nadolski <enadolski@suse.com>
Reviewed-by: Jeff Mahoney <jeffm@suse.com>
---
 fs/btrfs/backref.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 7699e16..e794b6e 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -26,6 +26,11 @@
 #include "delayed-ref.h"
 #include "locking.h"
 
+enum merge_mode {
+	MERGE_IDENTICAL_KEYS = 1,
+	MERGE_IDENTICAL_PARENTS,
+};
+
 /* Just an arbitrary number so we can be sure this happened */
 #define BACKREF_FOUND_SHARED 6
 
@@ -809,14 +814,12 @@ static int __add_missing_keys(struct btrfs_fs_info *fs_info,
 /*
  * merge backrefs and adjust counts accordingly
  *
- * mode = 1: merge identical keys, if key is set
  *    FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
  *           additionally, we could even add a key range for the blocks we
  *           looked into to merge even more (-> replace unresolved refs by those
  *           having a parent).
- * mode = 2: merge identical parents
  */
-static void __merge_refs(struct list_head *head, int mode)
+static void __merge_refs(struct list_head *head, enum merge_mode mode)
 {
 	struct __prelim_ref *pos1;
 
@@ -829,7 +832,7 @@ static void __merge_refs(struct list_head *head, int mode)
 
 			if (!ref_for_same_block(ref1, ref2))
 				continue;
-			if (mode == 1) {
+			if (mode == MERGE_IDENTICAL_KEYS) {
 				if (!ref1->parent && ref2->parent)
 					swap(ref1, ref2);
 			} else {
@@ -1374,7 +1377,7 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 	if (ret)
 		goto out;
 
-	__merge_refs(&prefs, 1);
+	__merge_refs(&prefs, MERGE_IDENTICAL_KEYS);
 
 	ret = __resolve_indirect_refs(fs_info, path, time_seq, &prefs,
 				      extent_item_pos, total_refs,
@@ -1382,7 +1385,7 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 	if (ret)
 		goto out;
 
-	__merge_refs(&prefs, 2);
+	__merge_refs(&prefs, MERGE_IDENTICAL_PARENTS);
 
 	while (!list_empty(&prefs)) {
 		ref = list_first_entry(&prefs, struct __prelim_ref, list);
-- 
2.10.2

[PATCH 2/2] btrfs: replace hardcoded value with SEQ_NONE macro

[ednadolski]

From: Edmund Nadolski <enadolski@suse.com>

Define the SEQ_NONE macro to replace (u64)-1 in places where said
value triggers a special-case ref search behavior.

Signed-off-by: Edmund Nadolski <enadolski@suse.com>
Reviewed-by: Jeff Mahoney <jeffm@suse.com>
---
 fs/btrfs/backref.c | 16 ++++++++--------
 fs/btrfs/backref.h |  2 ++
 fs/btrfs/qgroup.c  |  4 ++--
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index e794b6e..297de5b 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -538,7 +538,7 @@ static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 	 * slot==nritems. In that case, go to the next leaf before we continue.
 	 */
 	if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
-		if (time_seq == (u64)-1)
+		if (time_seq == SEQ_NONE)
 			ret = btrfs_next_leaf(root, path);
 		else
 			ret = btrfs_next_old_leaf(root, path, time_seq);
@@ -582,7 +582,7 @@ static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 			eie = NULL;
 		}
 next:
-		if (time_seq == (u64)-1)
+		if (time_seq == SEQ_NONE)
 			ret = btrfs_next_item(root, path);
 		else
 			ret = btrfs_next_old_item(root, path, time_seq);
@@ -634,7 +634,7 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 
 	if (path->search_commit_root)
 		root_level = btrfs_header_level(root->commit_root);
-	else if (time_seq == (u64)-1)
+	else if (time_seq == SEQ_NONE)
 		root_level = btrfs_header_level(root->node);
 	else
 		root_level = btrfs_old_root_level(root, time_seq);
@@ -645,7 +645,7 @@ static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 	}
 
 	path->lowest_level = level;
-	if (time_seq == (u64)-1)
+	if (time_seq == SEQ_NONE)
 		ret = btrfs_search_slot(NULL, root, &ref->key_for_search, path,
 					0, 0);
 	else
@@ -1199,7 +1199,7 @@ static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
  *
  * NOTE: This can return values > 0
  *
- * If time_seq is set to (u64)-1, it will not search delayed_refs, and behave
+ * If time_seq is set to SEQ_NONE, it will not search delayed_refs, and behave
  * much like trans == NULL case, the difference only lies in it will not
  * commit root.
  * The special case is for qgroup to search roots in commit_transaction().
@@ -1246,7 +1246,7 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 		path->skip_locking = 1;
 	}
 
-	if (time_seq == (u64)-1)
+	if (time_seq == SEQ_NONE)
 		path->skip_locking = 1;
 
 	/*
@@ -1276,9 +1276,9 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 
 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
 	if (trans && likely(trans->type != __TRANS_DUMMY) &&
-	    time_seq != (u64)-1) {
+	    time_seq != SEQ_NONE) {
 #else
-	if (trans && time_seq != (u64)-1) {
+	if (trans && time_seq != SEQ_NONE) {
 #endif
 		/*
 		 * look if there are updates for this ref queued and lock the
diff --git a/fs/btrfs/backref.h b/fs/btrfs/backref.h
index 9c41fba..20915a6 100644
--- a/fs/btrfs/backref.h
+++ b/fs/btrfs/backref.h
@@ -23,6 +23,8 @@
 #include "ulist.h"
 #include "extent_io.h"
 
+#define SEQ_NONE	((u64)-1)
+
 struct inode_fs_paths {
 	struct btrfs_path		*btrfs_path;
 	struct btrfs_root		*fs_root;
diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
index bb7e42f..76d84af 100644
--- a/fs/btrfs/qgroup.c
+++ b/fs/btrfs/qgroup.c
@@ -2077,12 +2077,12 @@ int btrfs_qgroup_account_extents(struct btrfs_trans_handle *trans,
 					goto cleanup;
 			}
 			/*
-			 * Use (u64)-1 as time_seq to do special search, which
+			 * Use SEQ_NONE as time_seq to do special search, which
 			 * doesn't lock tree or delayed_refs and search current
 			 * root. It's safe inside commit_transaction().
 			 */
 			ret = btrfs_find_all_roots(trans, fs_info,
-					record->bytenr, (u64)-1, &new_roots);
+					record->bytenr, SEQ_NONE, &new_roots);
 			if (ret < 0)
 				goto cleanup;
 			if (qgroup_to_skip) {
-- 
2.10.2

Re: [PATCH 2/2] btrfs: replace hardcoded value with SEQ_NONE macro

[David Sterba]

On Mon, Mar 13, 2017 at 02:32:04PM -0600, ednadolski@gmail.com wrote:
> From: Edmund Nadolski <enadolski@suse.com>
> 
> Define the SEQ_NONE macro to replace (u64)-1 in places where said
> value triggers a special-case ref search behavior.

> index 9c41fba..20915a6 100644
> --- a/fs/btrfs/backref.h
> +++ b/fs/btrfs/backref.h
> @@ -23,6 +23,8 @@
>  #include "ulist.h"
>  #include "extent_io.h"
>  
> +#define SEQ_NONE	((u64)-1)

Can you please move the definition to ctree.h, near line 660, where
seq_list and SEQ_LIST_INIT are defined, so thay're all grouped together?

Re: [PATCH 1/2] btrfs: provide enumeration for __merge_refs mode argument

[David Sterba]

On Mon, Mar 13, 2017 at 02:32:03PM -0600, ednadolski@gmail.com wrote:
> @@ -809,14 +814,12 @@ static int __add_missing_keys(struct btrfs_fs_info *fs_info,
>  /*
>   * merge backrefs and adjust counts accordingly
>   *
> - * mode = 1: merge identical keys, if key is set
>   *    FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
>   *           additionally, we could even add a key range for the blocks we
>   *           looked into to merge even more (-> replace unresolved refs by those
>   *           having a parent).

The 'FIXME' seems to refer to mode = 1, but now that you remove it, it's
not clear what it's referring to. Mentioning MERGE_IDENTICAL_KEYS would
be good.

Re: [PATCH 2/2] btrfs: replace hardcoded value with SEQ_NONE macro

[Qu Wenruo]

At 03/15/2017 10:38 PM, David Sterba wrote:
> On Mon, Mar 13, 2017 at 02:32:04PM -0600, ednadolski@gmail.com wrote:
>> From: Edmund Nadolski <enadolski@suse.com>
>>
>> Define the SEQ_NONE macro to replace (u64)-1 in places where said
>> value triggers a special-case ref search behavior.
>
>> index 9c41fba..20915a6 100644
>> --- a/fs/btrfs/backref.h
>> +++ b/fs/btrfs/backref.h
>> @@ -23,6 +23,8 @@
>>  #include "ulist.h"
>>  #include "extent_io.h"
>>
>> +#define SEQ_NONE	((u64)-1)

The naming of SEQ_NONE sounds not that good to me.

The (u64)-1 is to to info the backref walker to only search current 
root, and no need to worry about delayed_refs, since the caller (qgroup) 
will ensure that no delayed_ref will exist.

While the name SEQ_NONE seems a little like to 0, which is far from the 
original meaning.

What about SEQ_FINAL or SEQ_LAST?
Since the timing we use (u64)-1 is just before switching commit roots, 
it would be better for the naming to indicate that.

Thanks,
Qu



>
> Can you please move the definition to ctree.h, near line 660, where
> seq_list and SEQ_LIST_INIT are defined, so thay're all grouped together?
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>

Re: [PATCH 2/2] btrfs: replace hardcoded value with SEQ_NONE macro

[David Sterba]

On Thu, Mar 16, 2017 at 09:59:38AM +0800, Qu Wenruo wrote:
> 
> 
> At 03/15/2017 10:38 PM, David Sterba wrote:
> > On Mon, Mar 13, 2017 at 02:32:04PM -0600, ednadolski@gmail.com wrote:
> >> From: Edmund Nadolski <enadolski@suse.com>
> >>
> >> Define the SEQ_NONE macro to replace (u64)-1 in places where said
> >> value triggers a special-case ref search behavior.
> >
> >> index 9c41fba..20915a6 100644
> >> --- a/fs/btrfs/backref.h
> >> +++ b/fs/btrfs/backref.h
> >> @@ -23,6 +23,8 @@
> >>  #include "ulist.h"
> >>  #include "extent_io.h"
> >>
> >> +#define SEQ_NONE	((u64)-1)
> 
> The naming of SEQ_NONE sounds not that good to me.
> 
> The (u64)-1 is to to info the backref walker to only search current 
> root, and no need to worry about delayed_refs, since the caller (qgroup) 
> will ensure that no delayed_ref will exist.
> 
> While the name SEQ_NONE seems a little like to 0, which is far from the 
> original meaning.
> 
> What about SEQ_FINAL or SEQ_LAST?

SEQ_LAST sounds good to me.

> Since the timing we use (u64)-1 is just before switching commit roots, 
> it would be better for the naming to indicate that.

Yes, makes sense.