These TPs can help us monitor iomap content for dax reads and writes.
Signed-off-by: Liu Bo <bo.li....@oracle.com>
---
fs/btrfs/inode.c | 9 ++++
include/trace/events/btrfs.h | 106 +++++++++++++++++++++++++++++++++++++++++++
2 files changed, 115 insertions(+)
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 9851422..b5bee38 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -8067,6 +8067,8 @@ btrfs_get_blocks_dax_fault(struct inode *inode, u64
start, u64 len,
struct extent_map *em;
int ret = 0;
+ trace_btrfs_get_blocks_dax_entry(inode, start, len, create);
+
if (!create && start >= i_size_read(inode)) {
return 0;
}
@@ -8190,12 +8192,19 @@ btrfs_get_blocks_dax_fault(struct inode *inode, u64
start, u64 len,
map_block:
ret = btrfs_em_to_iomap(root->fs_info, start, len, em, create, iomap);
+ if (!ret) {
+ if (create)
+ trace_btrfs_iomap_alloc(inode, start, len, iomap);
+ else
+ trace_btrfs_iomap_found(inode, start, len, iomap);
+ }
out:
free_extent_map(em);
ASSERT(lockstart < lockend);
unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
&cached_state, GFP_NOFS);
+ trace_btrfs_get_blocks_dax_exit(inode, start, len, create);
return ret;
}
diff --git a/include/trace/events/btrfs.h b/include/trace/events/btrfs.h
index 0e04208..f7eb44f 100644
--- a/include/trace/events/btrfs.h
+++ b/include/trace/events/btrfs.h
@@ -7,6 +7,7 @@
#include <linux/writeback.h>
#include <linux/tracepoint.h>
#include <trace/events/mmflags.h>
+#include <linux/iomap.h>
struct btrfs_root;
struct btrfs_fs_info;
@@ -1471,6 +1472,111 @@ TRACE_EVENT(qgroup_update_counters,
__entry->cur_new_count)
);
+DECLARE_EVENT_CLASS(btrfs_iomap_class,
+
+ TP_PROTO(struct inode *inode, u64 offset, u64 len, struct iomap *iomap),
+
+ TP_ARGS(inode, offset, len, iomap),
+
+ TP_STRUCT__entry_btrfs(
+ __field( u64, ino )
+ __field( u64, isize )
+ __field( u64, disk_isize )
+ __field( u64, root_objectid )
+ __field( u64, offset )
+ __field( u64, len )
+ __field( u64, startoff )
+ __field( u64, blockstart )
+ __field( u64, blocklen )
+ __field( int, type )
+ ),
+
+ TP_fast_assign_btrfs(btrfs_sb(inode->i_sb),
+ __entry->ino = btrfs_ino(inode);
+ __entry->isize = i_size_read(inode);
+ __entry->disk_isize = BTRFS_I(inode)->disk_i_size;
+ __entry->root_objectid =
+ BTRFS_I(inode)->root->root_key.objectid;
+ __entry->offset = offset;
+ __entry->len = len;
+ __entry->startoff = iomap ? iomap->offset : 0;
+ __entry->blockstart = iomap ? (iomap->blkno << 9) : 0;
+ __entry->blocklen = iomap ? iomap->length : 0;
+ __entry->type = iomap ? iomap->type : 0;
+ ),
+
+ TP_printk_btrfs("root 0x%llx(%s) ino 0x%llx size 0x%llx "
+ "disk_isize 0x%llx offset 0x%llx len 0x%llx "
+ "startoff 0x%llx blockstart %llu blocklen 0x%llx "
+ " type %d",
+ show_root_type(__entry->root_objectid),
+ __entry->ino,
+ __entry->isize,
+ __entry->disk_isize,
+ __entry->offset,
+ __entry->len,
+ __entry->startoff,
+ __entry->blockstart,
+ __entry->blocklen,
+ __entry->type)
+);
+
+#define DEFINE_IOMAP_EVENT(name) \
+DEFINE_EVENT(btrfs_iomap_class, name, \
+ TP_PROTO(struct inode *inode, u64 offset, u64 len, \
+ struct iomap *iomap), \
+ TP_ARGS(inode, offset, len, iomap));
+
+DEFINE_IOMAP_EVENT(btrfs_iomap_alloc)
+DEFINE_IOMAP_EVENT(btrfs_iomap_found)
+
+DECLARE_EVENT_CLASS(btrfs_get_blocks_dax,
+
+ TP_PROTO(struct inode *inode, u64 offset, u64 len, int create),
+
+ TP_ARGS(inode, offset, len, create),
+
+ TP_STRUCT__entry_btrfs(
+ __field( u64, root_objectid )
+ __field( u64, ino )
+ __field( u64, isize )
+ __field( u64, disk_isize )
+ __field( u64, offset )
+ __field( u64, len )
+ __field( int, create )
+ ),
+
+ TP_fast_assign_btrfs(btrfs_sb(inode->i_sb),
+ __entry->root_objectid = BTRFS_I(inode)->root->objectid;
+ __entry->ino = btrfs_ino(inode);
+ __entry->isize = i_size_read(inode);
+ __entry->disk_isize = BTRFS_I(inode)->disk_i_size;
+ __entry->offset = offset;
+ __entry->len = len;
+ __entry->create = create;
+ ),
+
+ TP_printk_btrfs("root 0x%llx(%s) ino 0x%llx isize 0x%llx "
+ "disk_isize 0x%llx offset 0x%llx len 0x%llx create %d",
+ show_root_type(__entry->root_objectid),
+ __entry->ino,
+ __entry->isize,
+ __entry->disk_isize,
+ __entry->offset,
+ __entry->len,
+ __entry->create
+ )
+);
+
+#define DEFINE_GETBLOCKS_DAX_EVENT(name) \
+DEFINE_EVENT(btrfs_get_blocks_dax, name, \
+ TP_PROTO(struct inode *inode, u64 offset, u64 len, \
+ int create), \
+ TP_ARGS(inode, offset, len, create));
+
+DEFINE_GETBLOCKS_DAX_EVENT(btrfs_get_blocks_dax_entry)
+DEFINE_GETBLOCKS_DAX_EVENT(btrfs_get_blocks_dax_exit)
+
#endif /* _TRACE_BTRFS_H */
/* This part must be outside protection */
--
2.5.5
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
