bcachefs: better journal pipelining

Recently a severe performance regression was discovered, which bisected
to

  a6548c8b5eb5 bcachefs: Avoid flushing the journal in the discard path

It turns out the old behaviour, which issued excessive journal flushes,
worked around a performance issue where queueing delays would cause the
journal to not be able to write quickly enough and stall.

The journal flushes masked the issue because they periodically flushed
the device write cache, reducing write latency for non flushes.

This patch reworks the journalling code to allow more than one
(non-flush) write to be in flight at a time. With this patch, doing 4k
random writes and an iodepth of 128, we are now able to hit 560k iops to
a Samsung 970 EVO Plus - previously, we were stuck in the ~200k range.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

1 files changed, 7 insertions, 4 deletions

diff --git a/fs/bcachefs/journal_types.h b/fs/bcachefs/journal_types.h
index d75fbd3b1d34..3696aac3ccb7 100644
--- a/fs/bcachefs/journal_types.h
+++ b/fs/bcachefs/journal_types.h
@@ -34,10 +34,13 @@ struct journal_buf {
 	unsigned		disk_sectors;	/* maximum size entry could have been, if
 						   buf_size was bigger */
 	unsigned		u64s_reserved;
-	bool			noflush;	/* write has already been kicked off, and was noflush */
-	bool			must_flush;	/* something wants a flush */
-	bool			separate_flush;
-	bool			need_flush_to_write_buffer;
+	bool			noflush:1;	/* write has already been kicked off, and was noflush */
+	bool			must_flush:1;	/* something wants a flush */
+	bool			separate_flush:1;
+	bool			need_flush_to_write_buffer:1;
+	bool			write_started:1;
+	bool			write_allocated:1;
+	bool			write_done:1;
 	u8			idx;
 };