On 05/27/2017 02:08 AM, David Sterba wrote:
On Fri, May 26, 2017 at 03:45:00PM +0800, Anand Jain wrote:
This patch adds compression and decompression trace points for the
purpose of debugging.
Signed-off-by: Anand Jain <anand.jain@xxxxxxxxxx>
---
v2: no change
fs/btrfs/compression.c | 10 ++++++++++
include/trace/events/btrfs.h | 36 ++++++++++++++++++++++++++++++++++++
2 files changed, 46 insertions(+)
diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index e426a8f427b5..490590e62a2f 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -899,6 +899,10 @@ int btrfs_compress_pages(int type, struct address_space *mapping,
start, pages,
out_pages,
total_in, total_out);
+
+ trace_btrfs_encoder(1, 0, mapping->host, type, *total_in,
+ *total_out, start, ret);
+
free_workspace(type, workspace);
return ret;
}
@@ -927,6 +931,9 @@ static int btrfs_decompress_bio(struct compressed_bio *cb)
ret = btrfs_compress_op[type-1]->decompress_bio(workspace, cb);
+ trace_btrfs_encoder(0, 1, cb->inode, type,
+ cb->compressed_len, cb->len, cb->start, ret);
+
free_workspace(type, workspace);
return ret;
}
@@ -948,6 +955,9 @@ int btrfs_decompress(int type, unsigned char *data_in, struct page *dest_page,
dest_page, start_byte,
srclen, destlen);
+ trace_btrfs_encoder(0, 0, dest_page->mapping->host,
+ type, srclen, destlen, start_byte, ret);
+
free_workspace(type, workspace);
return ret;
}
diff --git a/include/trace/events/btrfs.h b/include/trace/events/btrfs.h
index e37973526153..1ebffcd005a1 100644
--- a/include/trace/events/btrfs.h
+++ b/include/trace/events/btrfs.h
@@ -1658,6 +1658,42 @@ TRACE_EVENT(qgroup_meta_reserve,
show_root_type(__entry->refroot), __entry->diff)
);
+TRACE_EVENT(btrfs_encoder,
So far the encoder is just compression/decompression. Having one
tracepoint for both operations makes more sense, but I don't like the
name.
I am ok with any suggestion.
In the long this will trace encryption as well so I was avoiding
compress specific term here. The other choice I had was
btrfs_tfm (transformer)
Also planning to rename compression.c to encoder.c (or anything
suggested), struct compressed_bio to encoder_bio (or anything
suggested). etc..
+
+ TP_PROTO(int encode, int bio, struct inode *inode, int type,
encode is confusing here, it seems to be an operation type
bio as an int is confusing, seems to describe the data source type, or
what 'pge' is supposed to mean
I had a version with the string passed, feel that I was better.
Will change it.
+ unsigned long bfr, unsigned long aft,
please do not abbreviate that much, this is not necessary and makes it
unreadable
ok will change it.
+ unsigned long start, int ret),
+
+ TP_ARGS(encode, bio, inode, type, bfr, aft, start, ret),
+
+ TP_STRUCT__entry_btrfs(
+ __field( int, encode)
+ __field( int, bio)
+ __field( unsigned long, i_ino)
+ __field( int, type)
+ __field( unsigned long, bfr)
+ __field( unsigned long, aft)
+ __field( unsigned long, start)
+ __field( int, ret)
+ ),
+
+ TP_fast_assign_btrfs(btrfs_sb(inode->i_sb),
+ __entry->encode = encode;
+ __entry->bio = bio;
+ __entry->i_ino = inode->i_ino;
+ __entry->type = type;
+ __entry->bfr = bfr;
+ __entry->aft = aft;
+ __entry->start = start;
+ __entry->ret = ret;
+ ),
+
+ TP_printk_btrfs("%s %s ino:%lu tfm:%d bfr:%lu aft:%lu start:%lu ret:%d",
please use "=" instead of ":"
I'll do that.
Thanks, Anand
+ __entry->encode ? "encode":"decode",
+ __entry->bio ? "bio":"pge", __entry->i_ino, __entry->type,
+ __entry->bfr, __entry->aft, __entry->start, __entry->ret)
+
+);
#endif /* _TRACE_BTRFS_H */
/* This part must be outside protection */
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
