Jens suggested that cloning the bio to issue a bare flush might be copying some flags we don't want, and at least allocating a new bio would be more correct. Worth a shot...
Signed-off-by: Kent Overstreet <[email protected]> --- On Tue, Sep 17, 2013 at 08:54:57PM +0000, Hubert Kempkensteffen wrote: > Kent Overstreet <kmo@...> writes: > Sorry, my english. > I have no raid and no LVM. There are both ordinary machines. > And I just want to speed up the data disk! > > > > > I think this bug requires something else in the stack to trigger, I'm > > trying to figure out what the common factor is. Can you give this patch a try? drivers/md/bcache/request.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c index afb9a99..c06840b 100644 --- a/drivers/md/bcache/request.c +++ b/drivers/md/bcache/request.c @@ -1053,19 +1053,21 @@ static void request_write(struct cached_dev *dc, struct search *s) trace_bcache_writethrough(s->orig_bio); closure_bio_submit(bio, cl, s->d); } else { + s->op.cache_bio = bio; trace_bcache_writeback(s->orig_bio); bch_writeback_add(dc, bio_sectors(bio)); - if (s->op.flush_journal) { + if (bio->bi_rw & REQ_FLUSH) { /* Also need to send a flush to the backing device */ - s->op.cache_bio = bio_clone_bioset(bio, GFP_NOIO, - dc->disk.bio_split); - - bio->bi_size = 0; - bio->bi_vcnt = 0; - closure_bio_submit(bio, cl, s->d); - } else { - s->op.cache_bio = bio; + struct bio *flush = bio_alloc_bioset(0, GFP_NOIO, + dc->disk.bio_split); + + flush->bi_rw = WRITE_FLUSH; + flush->bi_bdev = bio->bi_bdev; + flush->bi_end_io = request_endio; + flush->bi_private = cl; + + closure_bio_submit(flush, cl, s->d); } } out: -- 1.8.4.rc3 -- To unsubscribe from this list: send the line "unsubscribe linux-bcache" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
