Make the mq policy an alias for smq

Can you add in some documentation on how you tell which dm_cache policy is actually being used, and how to measure it, etc? It's a black box and some info would be nice.

Quoted 1 line

Quoted 1 line

Quoted 1 line

Quoted 1 line

"Joe" == Joe Thornber <[email protected]> writes:

Quoted 4 lines

Can you add in some documentation on how you tell which dm_cache policy is actually being used, and how to measure it, etc? It's a black box and some info would be nice.

smq seems to be performing better than the old mq policy in all situations, as well as using a quarter of the memory.

This patch makes 'mq' and alias for 'smq' when choosing a cache policy. The tunables that were present for the old mq are faked, and have no effect. mq should be considered deprecated now.

---
 drivers/md/Kconfig               |   11 +-
 drivers/md/Makefile              |    2 -
 drivers/md/dm-cache-policy-mq.c  | 1472 --------------------------------------
 drivers/md/dm-cache-policy-smq.c |   93 ++-
 drivers/md/dm-cache-policy.c     |    6 +-
 5 files changed, 97 insertions(+), 1487 deletions(-)
 delete mode 100644 drivers/md/dm-cache-policy-mq.c

drivers/md/Kconfig +4 -7

index b871928..9ea570b 100644
@@ -295,14 +295,11 @@ config DM_CACHE
          cleaned etc.  It supports writeback and writethrough modes.
 
 config DM_CACHE_MQ
-       tristate "MQ Cache Policy (EXPERIMENTAL)"
-       depends on DM_CACHE
-       default y
+       tristate "MQ Cache Policy (DEPRECATED)"
+       depends on DM_CACHE_SMQ
+       default n
        ---help---
-         A cache policy that uses a multiqueue ordered by recent hit
-         count to select which blocks should be promoted and demoted.
-         This is meant to be a general purpose policy.  It prioritises
-         reads over writes.
+         This policy is now an alias for the SMQ policy.
 
 config DM_CACHE_SMQ
        tristate "Stochastic MQ Cache Policy (EXPERIMENTAL)"

drivers/md/dm-cache-policy-smq.c +0 -0

index 1ffbeb1..8ad6b22 100644

drivers/md/dm-cache-policy-smq.c +18 -0

@@ -1716,6 +1781,15 @@ static struct dm_cache_policy_type smq_policy_type = {
 	.create = smq_create
 };
 
+static struct dm_cache_policy_type mq_policy_type = {
+	.name = "mq",
+	.version = {1, 4, 0},
+	.hint_size = 4,
+	.owner = THIS_MODULE,
+	.create = mq_create,
+	.real = &mq_policy_type
+};
+
 static struct dm_cache_policy_type default_policy_type = {
 	.name = "default",
 	.version = {1, 4, 0},

@@ -1735,9 +1809,17 @@ static int __init smq_init(void)
 		return -ENOMEM;
 	}
 
+	r = dm_cache_policy_register(&mq_policy_type);
+	if (r) {
+		DMERR("register failed %d", r);
+		dm_cache_policy_unregister(&smq_policy_type);
+		return -ENOMEM;
+	}
+
 	r = dm_cache_policy_register(&default_policy_type);
 	if (r) {
 		DMERR("register failed (as default) %d", r);
+		dm_cache_policy_unregister(&mq_policy_type);
 		dm_cache_policy_unregister(&smq_policy_type);
 		return -ENOMEM;
 	}

@@ -1748,6 +1830,7 @@ static int __init smq_init(void)
 static void __exit smq_exit(void)
 {
 	dm_cache_policy_unregister(&smq_policy_type);
+	dm_cache_policy_unregister(&mq_policy_type);
 	dm_cache_policy_unregister(&default_policy_type);
 }
 

drivers/md/dm-cache-policy.c +5 -1

index c1a3cee..829f3af 100644
@@ -62,7 +62,11 @@ static struct dm_cache_policy_type *get_policy(const char *name)
 	if (t)
 		return t;
 
-	request_module("dm-cache-%s", name);
+	/* hack; mq is in the smq module */
+	if (!strcmp(name, "mq"))
+		request_module("dm-cache-smq");
+	else
+		request_module("dm-cache-%s", name);
 
 	t = get_policy_once(name);
 	if (IS_ERR(t))

Signature · 1 line

2.5.0

Or is this some subtle compatability so that "mq" is always displayed in cache status output?

Shouldn't get_policy()'s call to get_policy_once() resolve "mq" to be "smq" if we just add this to the bottom of dm-cache-policy-smq?:

MODULE_ALIAS("dm-cache-mq");

Quoted 4 lines · Mike Snitzer

Shouldn't get_policy()'s call to get_policy_once() resolve "mq" to be "smq" if we just add this to the bottom of dm-cache-policy-smq?:

MODULE_ALIAS("dm-cache-mq");

Yes, you're right. See my fixup patch.

Quoted 7 lines · Joe Thornber

Quoted 4 lines · Mike Snitzer

Shouldn't get_policy()'s call to get_policy_once() resolve "mq" to be "smq" if we just add this to the bottom of dm-cache-policy-smq?:

MODULE_ALIAS("dm-cache-mq");

Yes, you're right. See my fixup patch.

Hmm, not seeing what you consider to be the latest fixed up version.. (nothing that sets: MODULE_ALIAS("dm-cache-mq") in smq.c)

But I've staged this in linux-next for 4.6, see:

https://git.kernel.org/cgit/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=for-next&id=8cbd0cb8cca67baa3e430fb6d636613fdc1cee1a