bcache: provide a switch to bypass all IO requests

euleros inclusion
category: feature
bugzilla: https://gitee.com/openeuler/kernel/issues/I4LOJ6
CVE: NA

------------------------------

provide a switch named read_bypass. If enbale, all IO requests will
bypass the cache. This option could be useful when we enable userspace
prefetch and the cache device is low capacity.
Signed-off-by: NLi Ruilin <liruilin4@huawei.com>
Reviewed-by: NLuan Jianhai <luanjianhai@huawei.com>
Reviewed-by: NPeng Junyi <pengjunyi1@huawei.com>
Acked-by: NXie Xiuqi <xiexiuqi@huawei.com>
Signed-off-by: NCheng Jian <cj.chengjian@huawei.com>
Reviewed-by: NGuangxing Deng <dengguangxing@huawei.com>
Reviewed-by: Nchao song <chao.song@huawei.com>
Reviewed-by: Nchao song <chao.song@huawei.com>
Signed-off-by: NZheng Zengkai <zhengzengkai@huawei.com>

Documentation/admin-guide/bcache.rst

@@ -441,6 +441,10 @@ sequential_cutoff
   most recent 128 IOs are tracked so sequential IO can be detected even when
   it isn't all done at once.
 
+read_bypass
+  If enbale, all IO will bypass the cache. This option could be useful when we
+  enable userspace prefetch and the cache device is low capacity.
+
 sequential_merge
   If non zero, bcache keeps a list of the last 128 requests submitted to compare
   against all new requests to determine which new requests are sequential

drivers/md/bcache/bcache.h

@@ -376,6 +376,8 @@ struct cached_dev {
 	unsigned char		writeback_percent;
 	unsigned int		writeback_delay;
 
+	unsigned int		read_bypass;
+
 	uint64_t		writeback_rate_target;
 	int64_t			writeback_rate_proportional;
 	int64_t			writeback_rate_integral;

drivers/md/bcache/request.c

@@ -852,7 +852,7 @@ static void cached_dev_read_done(struct closure *cl)
 	if (!s->prefetch)
 		bio_complete(s);
 
-	if (s->iop.bio &&
+	if (s->iop.bio && (!dc->read_bypass || s->prefetch) &&
 	    !test_bit(CACHE_SET_STOPPING, &s->iop.c->flags)) {
 		BUG_ON(!s->iop.replace);
 		closure_call(&s->iop.cl, bch_data_insert, NULL, cl);
@@ -897,12 +897,14 @@ static int cached_dev_cache_miss(struct btree *b, struct search *s,
 
 	s->cache_missed = 1;
 
-	if (s->cache_miss || s->iop.bypass) {
+	if (s->cache_miss || s->iop.bypass ||
+	    (dc->read_bypass && !s->prefetch)) {
 		miss = bio_next_split(bio, sectors, GFP_NOIO, &s->d->bio_split);
 		ret = miss == bio ? MAP_DONE : MAP_CONTINUE;
 		goto out_submit;
 	}
 
+	/* if called form do_readahead, no need to do this */
 	if (!(bio->bi_opf & REQ_RAHEAD) &&
 	    !(bio->bi_opf & (REQ_META|REQ_PRIO) ) &&
 	    s->iop.c->gc_stats.in_use < CUTOFF_CACHE_READA &&

drivers/md/bcache/super.c

@@ -1439,6 +1439,7 @@ static int cached_dev_init(struct cached_dev *dc, unsigned int block_size)
 	bch_cache_accounting_init(&dc->accounting, &dc->disk.cl);
 
 	dc->sequential_cutoff		= 4 << 20;
+	dc->read_bypass			= 0;
 
 	for (io = dc->io; io < dc->io + RECENT_IO; io++) {
 		list_add(&io->lru, &dc->io_lru);

drivers/md/bcache/sysfs.c

@@ -108,6 +108,7 @@ rw_attribute(congested_read_threshold_us);
 rw_attribute(congested_write_threshold_us);
 
 rw_attribute(sequential_cutoff);
+rw_attribute(read_bypass);
 rw_attribute(data_csum);
 rw_attribute(cache_mode);
 rw_attribute(readahead_cache_policy);
@@ -252,6 +253,7 @@ SHOW(__bch_cached_dev)
 	var_printf(partial_stripes_expensive,	"%u");
 
 	var_hprint(sequential_cutoff);
+	var_print(read_bypass);
 	var_hprint(readahead);
 
 	sysfs_print(running,		atomic_read(&dc->running));
@@ -346,6 +348,9 @@ STORE(__cached_dev)
 	sysfs_strtoul_clamp(sequential_cutoff,
 			    dc->sequential_cutoff,
 			    0, UINT_MAX);
+	sysfs_strtoul_clamp(read_bypass,
+			    dc->read_bypass,
+			    0, 1);
 	d_strtoi_h(readahead);
 
 	if (attr == &sysfs_clear_stats)
@@ -511,6 +516,7 @@ static struct attribute *bch_cached_dev_files[] = {
 	&sysfs_stripe_size,
 	&sysfs_partial_stripes_expensive,
 	&sysfs_sequential_cutoff,
+	&sysfs_read_bypass,
 	&sysfs_clear_stats,
 	&sysfs_running,
 	&sysfs_state,