btrfs: print-tree: output enhancement

PATCH In review Qu Wenruo 3 participants 6 messages Apr 11, 2018 – Apr 19, 2018 Btrfs file system

lore

Share & tools

lore

Qu Wenruo

author

2018-04-11 09:08 UTC #0 lore raw

PATCH 1/2 btrfs: print-tree: output enhancement

This patch enhance the following things:

- tree block header
* add generation and owner output for node and leaf
- node pointer generation output
- allow btrfs_print_tree() to not follow nodes
* just like btrfs-progs

Please note that, although function btrfs_print_tree() is not called by anyone right now, it's still a pretty useful function to debug kernel. So that function is still kept for later use.

Signed-off-by Qu Wenruo <[email protected]>

---
Fix title from "btrfs-progs" to "btrfs"

---
 fs/btrfs/print-tree.c | 25 +++++++++++++++----------
 fs/btrfs/print-tree.h |  2 +-
 2 files changed, 16 insertions(+), 11 deletions(-)

fs/btrfs/print-tree.c +15 ~~-10~~

index 4a8770485f77..9904cf741e1c 100644
@@ -202,9 +202,10 @@ void btrfs_print_leaf(struct extent_buffer *l)
 	fs_info = l->fs_info;
 	nr = btrfs_header_nritems(l);
 
-	btrfs_info(fs_info, "leaf %llu total ptrs %d free space %d",
-		   btrfs_header_bytenr(l), nr,
-		   btrfs_leaf_free_space(fs_info, l));
+	btrfs_info(fs_info,
+		   "leaf %llu gen %llu total ptrs %d free space %d owner %llu",
+		   btrfs_header_bytenr(l), btrfs_header_generation(l), nr,
+		   btrfs_leaf_free_space(fs_info, l), btrfs_header_owner(l));
 	for (i = 0 ; i < nr ; i++) {
 		item = btrfs_item_nr(i);
 		btrfs_item_key_to_cpu(l, &key, i);

@@ -338,7 +339,7 @@ void btrfs_print_leaf(struct extent_buffer *l)
 	}
 }
 
-void btrfs_print_tree(struct extent_buffer *c)
+void btrfs_print_tree(struct extent_buffer *c, bool follow)
 {
 	struct btrfs_fs_info *fs_info;
 	int i; u32 nr;

@@ -355,15 +356,19 @@ void btrfs_print_tree(struct extent_buffer *c)
 		return;
 	}
 	btrfs_info(fs_info,
-		   "node %llu level %d total ptrs %d free spc %u",
-		   btrfs_header_bytenr(c), level, nr,
-		   (u32)BTRFS_NODEPTRS_PER_BLOCK(fs_info) - nr);
+		   "node %llu level %d gen %llu total ptrs %d free spc %u owner %llu",
+		   btrfs_header_bytenr(c), level, btrfs_header_generation(c),
+		   nr, (u32)BTRFS_NODEPTRS_PER_BLOCK(fs_info) - nr,
+		   btrfs_header_owner(c));
 	for (i = 0; i < nr; i++) {
 		btrfs_node_key_to_cpu(c, &key, i);
-		pr_info("\tkey %d (%llu %u %llu) block %llu\n",
+		pr_info("\tkey %d (%llu %u %llu) block %llu gen %llu\n",
 		       i, key.objectid, key.type, key.offset,
-		       btrfs_node_blockptr(c, i));
+		       btrfs_node_blockptr(c, i),
+		       btrfs_node_ptr_generation(c, i));
 	}
+	if (!follow)
+		return;
 	for (i = 0; i < nr; i++) {
 		struct btrfs_key first_key;
 		struct extent_buffer *next;

@@ -385,7 +390,7 @@ void btrfs_print_tree(struct extent_buffer *c)
 		if (btrfs_header_level(next) !=
 		       level - 1)
 			BUG();
-		btrfs_print_tree(next);
+		btrfs_print_tree(next, follow);
 		free_extent_buffer(next);
 	}
 }

fs/btrfs/print-tree.h +1 -1

index 3afd508ed8c5..834f3c187f0d 100644
@@ -19,5 +19,5 @@
 #ifndef __PRINT_TREE_
 #define __PRINT_TREE_
 void btrfs_print_leaf(struct extent_buffer *l);
-void btrfs_print_tree(struct extent_buffer *c);
+void btrfs_print_tree(struct extent_buffer *c, bool follow);
 #endif

Signature · 1 line

2.17.0

Qu Wenruo

author

2018-04-11 09:08 UTC #1 lore raw

PATCH 2/2 btrfs: print-tree: Add locking status output for debug build

It's pretty handy if we can get debug output for locking status of an extent buffer, specially for race related debugging.

So add the following output for btrfs_print_tree() and btrfs_print_leaf():
- refs
- write_locks (as w:%u)
- read_locks (as r:%u)
- blocking_writers (as bw:%u)
- blocking_readers (as br:%u)
- spinning_writers (as sw:%u)
- spinning_readers (as sr:%u)
- lock_owner
- current->pid

Signed-off-by Qu Wenruo <[email protected]>

---
 fs/btrfs/print-tree.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

fs/btrfs/print-tree.c +22 -0

index 9904cf741e1c..7f0ada51e5fe 100644
@@ -179,6 +179,26 @@ static void print_uuid_item(struct extent_buffer *l, unsigned long offset,
 	}
 }
 
+/*
+ * Helper to output refs and locking status.
+ *
+ * Useful to debug race related problem
+ */
+static void print_eb_refs_lock(struct extent_buffer *eb)
+{
+#ifdef CONFIG_BTRFS_DEBUG
+	btrfs_info(eb->fs_info,
+"refs %u lock(w:%u r:%u bw:%u br:%u sw:%u sr:%u) lock_owner %u current %u",
+		   atomic_read(&eb->refs), atomic_read(&eb->write_locks),
+		   atomic_read(&eb->read_locks),
+		   atomic_read(&eb->blocking_writers),
+		   atomic_read(&eb->blocking_readers),
+		   atomic_read(&eb->spinning_writers),
+		   atomic_read(&eb->spinning_readers),
+		   eb->lock_owner, current->pid);
+#endif
+}
+
 void btrfs_print_leaf(struct extent_buffer *l)
 {
 	struct btrfs_fs_info *fs_info;

@@ -206,6 +226,7 @@ void btrfs_print_leaf(struct extent_buffer *l)
 		   "leaf %llu gen %llu total ptrs %d free space %d owner %llu",
 		   btrfs_header_bytenr(l), btrfs_header_generation(l), nr,
 		   btrfs_leaf_free_space(fs_info, l), btrfs_header_owner(l));
+	print_eb_refs_lock(l);
 	for (i = 0 ; i < nr ; i++) {
 		item = btrfs_item_nr(i);
 		btrfs_item_key_to_cpu(l, &key, i);

@@ -360,6 +381,7 @@ void btrfs_print_tree(struct extent_buffer *c, bool follow)
 		   btrfs_header_bytenr(c), level, btrfs_header_generation(c),
 		   nr, (u32)BTRFS_NODEPTRS_PER_BLOCK(fs_info) - nr,
 		   btrfs_header_owner(c));
+	print_eb_refs_lock(c);
 	for (i = 0; i < nr; i++) {
 		btrfs_node_key_to_cpu(c, &key, i);
 		pr_info("\tkey %d (%llu %u %llu) block %llu gen %llu\n",

Signature · 1 line

2.17.0

David Sterba

2018-04-19 13:04 UTC #2 lore raw

Re: [PATCH 2/2] btrfs: print-tree: Add locking status output for debug build

Quoted 14 lines · Qu Wenruo

It's pretty handy if we can get debug output for locking status of an extent buffer, specially for race related debugging.
So add the following output for btrfs_print_tree() and btrfs_print_leaf():
- refs
- write_locks (as w:%u)
- read_locks (as r:%u)
- blocking_writers (as bw:%u)
- blocking_readers (as br:%u)
- spinning_writers (as sw:%u)
- spinning_readers (as sr:%u)
- lock_owner
- current->pid

Useful, but atomic is 'int' (%d), and pid_t is also int in disguise (maybe there's a special printk specifier for that).

David Sterba

Reviewed-by on v1 1/2

2018-04-19 12:51 UTC #3 lore raw

In reply to Qu Wenruo

Quoted 11 lines · Qu Wenruo

This patch enhance the following things:
- tree block header
* add generation and owner output for node and leaf
- node pointer generation output
- allow btrfs_print_tree() to not follow nodes
* just like btrfs-progs
Please note that, although function btrfs_print_tree() is not called by anyone right now, it's still a pretty useful function to debug kernel. So that function is still kept for later use.

It will stay of course, code deletionists have no chance removing useful debugging stuff.

Reviewed-by David Sterba <[email protected]>

Some time ago I improved the progs output of print-tree, this would be also good to port to kernel and we can then keep the files in 1:1.

Qu Wenruo

2018-04-19 12:57 UTC #4 lore raw

Quoted 13 lines · David Sterba

Quoted 11 lines · Qu Wenruo
This patch enhance the following things:
- tree block header
* add generation and owner output for node and leaf
- node pointer generation output
- allow btrfs_print_tree() to not follow nodes
* just like btrfs-progs
Please note that, although function btrfs_print_tree() is not called by anyone right now, it's still a pretty useful function to debug kernel. So that function is still kept for later use.

Quoted 7 lines

It will stay of course, code deletionists have no chance removing useful debugging stuff.
Reviewed-by David Sterba <[email protected]>
Some time ago I improved the progs output of print-tree, this would be also good to port to kernel and we can then keep the files in 1:1.

Definitely!
The current kernel print-tree is just better than no output, far from human readable output.
I can't wait to port btrfs-progs print-tree to kernel, but that's over 1000 lines for debug facility.

If that's OK, I could start the port any time.

Thanks,
Qu

Quoted 5 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

David Sterba

2018-04-19 13:14 UTC #5 lore raw

Quoted 24 lines · Qu Wenruo

Quoted 13 lines · David Sterba
Quoted 11 lines · Qu Wenruo
This patch enhance the following things:
- tree block header
* add generation and owner output for node and leaf
- node pointer generation output
- allow btrfs_print_tree() to not follow nodes
* just like btrfs-progs
Please note that, although function btrfs_print_tree() is not called by anyone right now, it's still a pretty useful function to debug kernel. So that function is still kept for later use.
Quoted 7 lines
It will stay of course, code deletionists have no chance removing useful debugging stuff.
Reviewed-by David Sterba <[email protected]>
Some time ago I improved the progs output of print-tree, this would be also good to port to kernel and we can then keep the files in 1:1.

Quoted 7 lines

Definitely!
The current kernel print-tree is just better than no output, far from human readable output.
I can't wait to port btrfs-progs print-tree to kernel, but that's over 1000 lines for debug facility.
If that's OK, I could start the port any time.

As print-tree is independent, you can do that incrementally, similar to the progs patches that factor out the per-key cases into functions. Anything you'd have we can merge rightaway, these are easy patches that can serve as a "relaxation" and relief from the difficult ones.