[RFC PATCH 1/2] vfs: tracepoints for buffer flag updates

[RFC PATCH 1/2] vfs: tracepoints for buffer flag updates

[Aneesh Kumar K.V]

We collect the device details so that we can filter with device value

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/buffer.c                        |    2 +
 include/linux/buffer_head.h        |   11 +++-
 include/trace/events/buffer_head.h |   95 ++++++++++++++++++++++++++++++++++++
 3 files changed, 106 insertions(+), 2 deletions(-)
 create mode 100644 include/trace/events/buffer_head.h

diff --git a/fs/buffer.c b/fs/buffer.c
index 6fa5302..6f441aa 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -41,6 +41,8 @@
 #include <linux/bitops.h>
 #include <linux/mpage.h>
 #include <linux/bit_spinlock.h>
+#define CREATE_TRACE_POINTS
+#include <trace/events/buffer_head.h>
 
 static int fsync_buffers_list(spinlock_t *lock, struct list_head *list);
 
diff --git a/include/linux/buffer_head.h b/include/linux/buffer_head.h
index 16ed028..1780701 100644
--- a/include/linux/buffer_head.h
+++ b/include/linux/buffer_head.h
@@ -13,6 +13,7 @@
 #include <linux/pagemap.h>
 #include <linux/wait.h>
 #include <asm/atomic.h>
+#include <trace/events/buffer_head.h>
 
 #ifdef CONFIG_BLOCK
 
@@ -84,10 +85,12 @@ struct buffer_head {
 static inline void set_buffer_##name(struct buffer_head *bh)		\
 {									\
 	set_bit(BH_##bit, &(bh)->b_state);				\
+	trace_set_buffer_##name(bh);					\
 }									\
 static inline void clear_buffer_##name(struct buffer_head *bh)		\
 {									\
 	clear_bit(BH_##bit, &(bh)->b_state);				\
+	trace_clear_buffer_##name(bh);					\
 }									\
 static inline int buffer_##name(const struct buffer_head *bh)		\
 {									\
@@ -100,11 +103,15 @@ static inline int buffer_##name(const struct buffer_head *bh)		\
 #define TAS_BUFFER_FNS(bit, name)					\
 static inline int test_set_buffer_##name(struct buffer_head *bh)	\
 {									\
-	return test_and_set_bit(BH_##bit, &(bh)->b_state);		\
+	int ret = test_and_set_bit(BH_##bit, &(bh)->b_state);		\
+	trace_set_buffer_##name(bh);					\
+	return ret;							\
 }									\
 static inline int test_clear_buffer_##name(struct buffer_head *bh)	\
 {									\
-	return test_and_clear_bit(BH_##bit, &(bh)->b_state);		\
+	int ret = test_and_clear_bit(BH_##bit, &(bh)->b_state);		\
+	trace_clear_buffer_##name(bh);					\
+	return ret;							\
 }									\
 
 /*
diff --git a/include/trace/events/buffer_head.h b/include/trace/events/buffer_head.h
new file mode 100644
index 0000000..07605b5
--- /dev/null
+++ b/include/trace/events/buffer_head.h
@@ -0,0 +1,95 @@
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM buffer_head
+
+#if !defined(_TRACE_BUFFER_HEAD_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_BUFFER_HEAD_H
+
+#include <linux/tracepoint.h>
+struct buffer_head;
+extern const char *jbd2_dev_to_name(dev_t device);
+
+
+#define BUFFER_HEAD_TRACE_EVENT(name)						\
+TRACE_EVENT(set_buffer_##name,							\
+	TP_PROTO(struct buffer_head *bh),					\
+										\
+	TP_ARGS(bh),								\
+										\
+	TP_STRUCT__entry(							\
+		__field(	void *,   buffer		)		\
+		__field(	sector_t, block			)		\
+		__field(	unsigned long,	state		)		\
+		__field(	dev_t,    dev			)		\
+	),									\
+										\
+	TP_fast_assign(								\
+		__entry->buffer = bh;						\
+		__entry->block	= bh->b_blocknr;				\
+		__entry->state	= bh->b_state;					\
+		if (bh->b_bdev)							\
+			__entry->dev	= disk_devt(bh->b_bdev->bd_disk);	\
+		else								\
+			__entry->dev	= 0;					\
+	),									\
+										\
+	TP_printk("bh = %p device = %s block = %llu new state = 0x%lx",		\
+			__entry->buffer, jbd2_dev_to_name(__entry->dev),	\
+			(unsigned long long) __entry->block,			\
+			  __entry->state)					\
+);										\
+										\
+TRACE_EVENT(clear_buffer_##name,						\
+	TP_PROTO(struct buffer_head *bh),					\
+										\
+	TP_ARGS(bh),								\
+										\
+	TP_STRUCT__entry(							\
+		__field(	void *,   buffer		)		\
+		__field(	sector_t, block			)		\
+		__field(	unsigned long,	  state		)		\
+		__field(	dev_t,    dev			)		\
+	),									\
+										\
+	TP_fast_assign(								\
+		__entry->buffer = bh;						\
+		__entry->block	= bh->b_blocknr;				\
+		__entry->state	= bh->b_state;					\
+		if (bh->b_bdev)							\
+			__entry->dev	= disk_devt(bh->b_bdev->bd_disk);	\
+		else								\
+			__entry->dev	= 0;					\
+	),									\
+										\
+	TP_printk("bh = %p device = %s block = %llu new state = 0x%lx",		\
+			__entry->buffer, jbd2_dev_to_name(__entry->dev),	\
+			(unsigned long long) __entry->block,			\
+			  __entry->state)					\
+)
+
+BUFFER_HEAD_TRACE_EVENT(uptodate);
+BUFFER_HEAD_TRACE_EVENT(dirty);
+BUFFER_HEAD_TRACE_EVENT(locked);
+BUFFER_HEAD_TRACE_EVENT(req);
+BUFFER_HEAD_TRACE_EVENT(mapped);
+BUFFER_HEAD_TRACE_EVENT(new);
+BUFFER_HEAD_TRACE_EVENT(async_read);
+BUFFER_HEAD_TRACE_EVENT(async_write);
+BUFFER_HEAD_TRACE_EVENT(delay);
+BUFFER_HEAD_TRACE_EVENT(boundary);
+BUFFER_HEAD_TRACE_EVENT(write_io_error);
+BUFFER_HEAD_TRACE_EVENT(ordered);
+BUFFER_HEAD_TRACE_EVENT(eopnotsupp);
+BUFFER_HEAD_TRACE_EVENT(unwritten);
+
+/* jbd */
+BUFFER_HEAD_TRACE_EVENT(jbd);
+BUFFER_HEAD_TRACE_EVENT(jwrite);
+BUFFER_HEAD_TRACE_EVENT(jbddirty);
+BUFFER_HEAD_TRACE_EVENT(revoked);
+BUFFER_HEAD_TRACE_EVENT(revokevalid);
+BUFFER_HEAD_TRACE_EVENT(freed);
+
+#endif /* _TRACE_BUFFER_HEAD_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>
-- 
1.6.5.2.74.g610f9

[RFC PATCH 2/2] ext[34]: Make sure we initialize the b_bdev field of dummy buffer_heads

[Aneesh Kumar K.V]

buffer_head flag update trace point look at the b_bdev field to track
the backing device. For dummy buffer_heads used we expect this to be
NULL.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/buffer.c       |    1 +
 fs/ext3/inode.c   |    1 +
 fs/ext4/inode.c   |    1 +
 fs/jbd/journal.c  |    1 +
 fs/jbd2/journal.c |    2 ++
 fs/mpage.c        |    1 +
 6 files changed, 7 insertions(+), 0 deletions(-)

diff --git a/fs/buffer.c b/fs/buffer.c
index 6f441aa..5c60fa3 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -2923,6 +2923,7 @@ sector_t generic_block_bmap(struct address_space *mapping, sector_t block,
 	struct inode *inode = mapping->host;
 	tmp.b_state = 0;
 	tmp.b_blocknr = 0;
+	tmp.b_bdev = NULL;
 	tmp.b_size = 1 << inode->i_blkbits;
 	get_block(inode, block, &tmp, 0);
 	return tmp.b_blocknr;
diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c
index acf1b14..c7508aa 100644
--- a/fs/ext3/inode.c
+++ b/fs/ext3/inode.c
@@ -1007,6 +1007,7 @@ struct buffer_head *ext3_getblk(handle_t *handle, struct inode *inode,
 
 	dummy.b_state = 0;
 	dummy.b_blocknr = -1000;
+	dummy.b_bdev = NULL;
 	buffer_trace_init(&dummy.b_history);
 	err = ext3_get_blocks_handle(handle, inode, block, 1,
 					&dummy, create);
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 5c5bc5d..dcb362d 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1392,6 +1392,7 @@ struct buffer_head *ext4_getblk(handle_t *handle, struct inode *inode,
 
 	dummy.b_state = 0;
 	dummy.b_blocknr = -1000;
+	dummy.b_bdev = NULL;
 	buffer_trace_init(&dummy.b_history);
 	if (create)
 		flags |= EXT4_GET_BLOCKS_CREATE;
diff --git a/fs/jbd/journal.c b/fs/jbd/journal.c
index bd3c073..1572a08 100644
--- a/fs/jbd/journal.c
+++ b/fs/jbd/journal.c
@@ -303,6 +303,7 @@ int journal_write_metadata_buffer(transaction_t *transaction,
 	new_bh = alloc_buffer_head(GFP_NOFS|__GFP_NOFAIL);
 	/* keep subsequent assertions sane */
 	new_bh->b_state = 0;
+	new_bh->b_bdev = NULL;
 	init_buffer(new_bh, NULL, NULL);
 	atomic_set(&new_bh->b_count, 1);
 	new_jh = journal_add_journal_head(new_bh);	/* This sleeps */
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index b0ab521..f18af08 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -2250,6 +2250,8 @@ const char *jbd2_dev_to_name(dev_t device)
 	struct block_device *bd;
 	static struct devname_cache *new_dev;
 
+	if (!device)
+		return "NULL";
 	rcu_read_lock();
 	if (devcache[i] && devcache[i]->device == device) {
 		ret = devcache[i]->devname;
diff --git a/fs/mpage.c b/fs/mpage.c
index 42381bd..eb72426 100644
--- a/fs/mpage.c
+++ b/fs/mpage.c
@@ -381,6 +381,7 @@ mpage_readpages(struct address_space *mapping, struct list_head *pages,
 
 	map_bh.b_state = 0;
 	map_bh.b_size = 0;
+	map_bh.b_bdev = NULL;
 	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
 		struct page *page = list_entry(pages->prev, struct page, lru);
 
-- 
1.6.5.2.74.g610f9

Re: [RFC PATCH 1/2] vfs: tracepoints for buffer flag updates

[Aneesh Kumar K.V]

This patchset when enabled with stacktrace will give output like the below

              dd-1905  [003]    26.445155: set_buffer_uptodate: bh = df5c28c0 device = vda block = 34817 new state= 0x23
              dd-1905  [003]    26.445156: <stack trace>
 => set_buffer_uptodate
 => __block_commit_write
 => block_write_end
 => ext3_writeback_write_end
 => generic_file_buffered_write
 => __generic_file_aio_write
 => generic_file_aio_write
 => do_sync_write
              dd-1905  [003]    26.445168: clear_buffer_new: bh = df5c28c0 device = vda block = 34817 new state = 0x23
              dd-1905  [003]    26.445169: <stack trace>
 => clear_buffer_new
 => __block_commit_write
 => block_write_end
 => ext3_writeback_write_end
 => generic_file_buffered_write
 => __generic_file_aio_write
 => generic_file_aio_write
 => do_sync_write
              dd-1905  [003]    26.445182: set_buffer_revokevalid: bh = df406540 device = vda block = 32835 new state= 0x310029
              dd-1905  [003]    26.445183: <stack trace>
 => journal_cancel_revoke
 => do_get_write_access
 => journal_get_write_access
 => __ext3_journal_get_write_access
 => ext3_reserve_inode_write
 => ext3_mark_inode_dirty
 => ext3_dirty_inode
 => __mark_inode_dirty
              dd-1905  [003]    26.445197: clear_buffer_revoked: bh = df406540 device = vda block = 32835 new state = 0x310029
              dd-1905  [003]    26.445198: <stack trace>
 => journal_cancel_revoke
 => do_get_write_access
 => journal_get_write_access
 => __ext3_journal_get_write_access
 => ext3_reserve_inode_write
 => ext3_mark_inode_dirty
 => ext3_dirty_inode
 => __mark_inode_dirty


-aneesh