Skip to contentSkip to conversation

misc changes for rtrs

PATCH In review Guoqing Jiang 4 participants 9 messages Sep 8, 2022 – Sep 20, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

v2 cover misc changes for rtrs

Hi,

Pls review the three patches.

Thanks,
Guoqing

V2 changes:
1. fix compile warnings in the third patch
2. collect tag from Haris, thanks!

Guoqing Jiang (3):
RDMA/rtrs: Update comments for MAX_SESS_QUEUE_DEPTH RDMA/rtrs-clt: Break the loop once one path is connected RDMA/rtrs-clt: Kill xchg_paths

 drivers/infiniband/ulp/rtrs/rtrs-clt.c | 19 ++++++-------------
 drivers/infiniband/ulp/rtrs/rtrs-pri.h |  7 +++----
 2 files changed, 9 insertions(+), 17 deletions(-)
Signature · 1 line
2.31.1
#1 lore raw

PATCH v2 1/3 RDMA/rtrs: Update comments for MAX_SESS_QUEUE_DEPTH

The maximum queue_depth should be 65535 per check_module_params, also update other relevant comments.

---
 drivers/infiniband/ulp/rtrs/rtrs-pri.h | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/infiniband/ulp/rtrs/rtrs-pri.h +3 -4
index ac0df734eba8..a2420eecaf5a 100644
@@ -26,11 +26,10 @@
 /*
  * Max IB immediate data size is 2^28 (MAX_IMM_PAYL_BITS)
  * and the minimum chunk size is 4096 (2^12).
- * So the maximum sess_queue_depth is 65536 (2^16) in theory.
- * But mempool_create, create_qp and ib_post_send fail with
- * "cannot allocate memory" error if sess_queue_depth is too big.
+ * So the maximum sess_queue_depth is 65535 (2^16 - 1) in theory
+ * since queue_depth in rtrs_msg_conn_rsp is defined as le16.
  * Therefore the pratical max value of sess_queue_depth is
- * somewhere between 1 and 65534 and it depends on the system.
+ * somewhere between 1 and 65535 and it depends on the system.
  */
 #define MAX_SESS_QUEUE_DEPTH 65535
 
Signature · 1 line
2.31.1
#2 lore raw

PATCH v2 2/3 RDMA/rtrs-clt: Break the loop once one path is connected

In reply to Guoqing Jiang

No need to iterate all paths after find one connected path.

---
 drivers/infiniband/ulp/rtrs/rtrs-clt.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/infiniband/ulp/rtrs/rtrs-clt.c +4 -1
index 5219bb10777a..c29eccdb4fd2 100644
@@ -54,7 +54,10 @@ static inline bool rtrs_clt_is_connected(const struct rtrs_clt_sess *clt)
 
 	rcu_read_lock();
 	list_for_each_entry_rcu(clt_path, &clt->paths_list, s.entry)
-		connected |= READ_ONCE(clt_path->state) == RTRS_CLT_CONNECTED;
+		if (READ_ONCE(clt_path->state) == RTRS_CLT_CONNECTED) {
+			connected = true;
+			break;
+		}
 	rcu_read_unlock();
 
 	return connected;
Signature · 1 line
2.31.1
#3 lore raw

PATCH v2 3/3 RDMA/rtrs-clt: Kill xchg_paths

In reply to Guoqing Jiang

Let's call try_cmpxchg directly for the same purpose.

---
 drivers/infiniband/ulp/rtrs/rtrs-clt.c | 14 ++------------
 1 file changed, 2 insertions(+), 12 deletions(-)
drivers/infiniband/ulp/rtrs/rtrs-clt.c +2 -12
index c29eccdb4fd2..2428bf5d07e3 100644
@@ -2220,17 +2220,6 @@ static void rtrs_clt_stop_and_destroy_conns(struct rtrs_clt_path *clt_path)
 	}
 }
 
-static inline bool xchg_paths(struct rtrs_clt_path __rcu **rcu_ppcpu_path,
-			      struct rtrs_clt_path *clt_path,
-			      struct rtrs_clt_path *next)
-{
-	struct rtrs_clt_path **ppcpu_path;
-
-	/* Call cmpxchg() without sparse warnings */
-	ppcpu_path = (typeof(ppcpu_path))rcu_ppcpu_path;
-	return clt_path == cmpxchg(ppcpu_path, clt_path, next);
-}
-
 static void rtrs_clt_remove_path_from_arr(struct rtrs_clt_path *clt_path)
 {
 	struct rtrs_clt_sess *clt = clt_path->clt;
@@ -2305,7 +2294,8 @@ static void rtrs_clt_remove_path_from_arr(struct rtrs_clt_path *clt_path)
 		 * We race with IO code path, which also changes pointer,
 		 * thus we have to be careful not to overwrite it.
 		 */
-		if (xchg_paths(ppcpu_path, clt_path, next))
+		if (try_cmpxchg((struct rtrs_clt_path **)ppcpu_path,
+				&clt_path, next))
 			/*
 			 * @ppcpu_path was successfully replaced with @next,
 			 * that means that someone could also pick up the
Signature · 1 line
2.31.1
maintainer
#4 lore raw

Re: [PATCH V2 3/3] RDMA/rtrs-clt: Kill xchg_paths

Quoted 6 lines · Guoqing Jiang

Let's call try_cmpxchg directly for the same purpose.

I am not sure whats the correct way of using this. But technically, this change was NOT done due to a report from the "kernel test robot". It only pointed out the problem in the original patch. To the branch maintainers, if its okay to keep this in this scenario, then ignore this comment.

Thanks.

Quoted 39 lines
---
 drivers/infiniband/ulp/rtrs/rtrs-clt.c | 14 ++------------
 1 file changed, 2 insertions(+), 12 deletions(-)
drivers/infiniband/ulp/rtrs/rtrs-clt.c +2 -13
index c29eccdb4fd2..2428bf5d07e3 100644
@@ -2220,17 +2220,6 @@ static void rtrs_clt_stop_and_destroy_conns(struct rtrs_clt_path *clt_path)
        }
 }
 
-static inline bool xchg_paths(struct rtrs_clt_path __rcu **rcu_ppcpu_path,
-                             struct rtrs_clt_path *clt_path,
-                             struct rtrs_clt_path *next)
-{
-       struct rtrs_clt_path **ppcpu_path;
-
-       /* Call cmpxchg() without sparse warnings */
-       ppcpu_path = (typeof(ppcpu_path))rcu_ppcpu_path;
-       return clt_path == cmpxchg(ppcpu_path, clt_path, next);
-}
-
 static void rtrs_clt_remove_path_from_arr(struct rtrs_clt_path *clt_path)
 {
        struct rtrs_clt_sess *clt = clt_path->clt;
@@ -2305,7 +2294,8 @@ static void rtrs_clt_remove_path_from_arr(struct rtrs_clt_path *clt_path)
                 * We race with IO code path, which also changes pointer,
                 * thus we have to be careful not to overwrite it.
                 */
-               if (xchg_paths(ppcpu_path, clt_path, next))
+               if (try_cmpxchg((struct rtrs_clt_path **)ppcpu_path,
+                               &clt_path, next))
                        /*
                         * @ppcpu_path was successfully replaced with @next,
                         * that means that someone could also pick up the
--

2.31.1

#5 lore raw

Re: [PATCH V2 3/3] RDMA/rtrs-clt: Kill xchg_paths

Quoted 11 lines · Haris Iqbal
Quoted 5 lines · Guoqing Jiang

Let's call try_cmpxchg directly for the same purpose.

I am not sure whats the correct way of using this. But technically, this change was NOT done due to a report from the "kernel test robot". It only pointed out the problem in the original patch. To the branch maintainers, if its okay to keep this in this scenario, then ignore this comment.

Me either, just want to give credit to lkp for previous report, but not sure if there is a better tag.

Thanks,
Guoqing

#6 lore raw

Re: [PATCH V2 3/3] RDMA/rtrs-clt: Kill xchg_paths

Looks good.

As clt->pcpu_path is of type "struct rtrs_clt_path __rcu **ppcpu_path", using without typecasting with cmpxchg() would fetch sparse warning.

I still believe we can use rcu_replace_pointer() instead of cmpxchg(), anyway we are going to allow RCU grace period using synchronize_rcu().

Best Regards,
Santosh

Quoted 21 lines · Guoqing Jiang
Quoted 11 lines · Haris Iqbal
Quoted 5 lines · Guoqing Jiang

Let's call try_cmpxchg directly for the same purpose.

I am not sure whats the correct way of using this. But technically, this change was NOT done due to a report from the "kernel test robot". It only pointed out the problem in the original patch. To the branch maintainers, if its okay to keep this in this scenario, then ignore this comment.

Me either, just want to give credit to lkp for previous report, but not sure if there is a better tag.

Thanks,
Guoqing

#7 lore raw
In reply to Guoqing Jiang
Quoted 7 lines · Guoqing Jiang

Hi,

Pls review the three patches.

Thanks,
Guoqing

Quoted 6 lines

V2 changes:
1. fix compile warnings in the third patch
2. collect tag from Haris, thanks!

Guoqing Jiang (3):
RDMA/rtrs: Update comments for MAX_SESS_QUEUE_DEPTH

This patch doesn't apply.

Quoted 10 lines

RDMA/rtrs-clt: Break the loop once one path is connected RDMA/rtrs-clt: Kill xchg_paths

 drivers/infiniband/ulp/rtrs/rtrs-clt.c | 19 ++++++-------------
 drivers/infiniband/ulp/rtrs/rtrs-pri.h |  7 +++----
 2 files changed, 9 insertions(+), 17 deletions(-)
Signature · 1 line
2.31.1
#8 lore raw
Quoted 8 lines · Leon Romanovsky
Quoted 7 lines · Guoqing Jiang

Hi,

Pls review the three patches.

Thanks,
Guoqing

Quoted 8 lines
Quoted 6 lines

V2 changes:
1. fix compile warnings in the third patch
2. collect tag from Haris, thanks!

Guoqing Jiang (3):
RDMA/rtrs: Update comments for MAX_SESS_QUEUE_DEPTH

This patch doesn't apply.

Actually, everything was already applied by me along time ago.

Quoted 11 lines
Quoted 11 lines

RDMA/rtrs-clt: Break the loop once one path is connected RDMA/rtrs-clt: Kill xchg_paths

 drivers/infiniband/ulp/rtrs/rtrs-clt.c | 19 ++++++-------------
 drivers/infiniband/ulp/rtrs/rtrs-pri.h |  7 +++----
 2 files changed, 9 insertions(+), 17 deletions(-)
Signature · 1 line
2.31.1