Skip to contentSkip to conversation

RDMA/rxe: Add function name to the logs

PATCH In review Zhu Yanjun 3 participants 4 messages Apr 18, 2023 – Apr 28, 2023

Subsystems (2)
lore
Share & tools
lore

Zhu Yanjun

author
#0 lore raw

PATCH 1/1 RDMA/rxe: Add function name to the logs

From: Zhu Yanjun <[email protected]>

Add the function names to the pr_ logs. As such, if some bugs occur, with function names, it is easy to locate the bugs.

---
 drivers/infiniband/sw/rxe/rxe.h       |  2 +-
 drivers/infiniband/sw/rxe/rxe_queue.h | 16 ++++------------
 2 files changed, 5 insertions(+), 13 deletions(-)
drivers/infiniband/sw/rxe/rxe.h +1 -1
index d33dd6cf83d3..f897d0ac1216 100644
@@ -10,7 +10,7 @@
 #ifdef pr_fmt
 #undef pr_fmt
 #endif
-#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#define pr_fmt(fmt) KBUILD_MODNAME ": %s: " fmt, __func__
 
 #include <linux/skbuff.h>
 
drivers/infiniband/sw/rxe/rxe_queue.h +4 -12
index c711cb98b949..11bba08ca5f9 100644
@@ -184,9 +184,7 @@ static inline void queue_advance_producer(struct rxe_queue *q,
 	switch (type) {
 	case QUEUE_TYPE_FROM_CLIENT:
 		/* used by rxe, client owns the index */
-		if (WARN_ON(1))
-			pr_warn("%s: attempt to advance client index\n",
-				__func__);
+		WARN_ON(1);
 		break;
 	case QUEUE_TYPE_TO_CLIENT:
 		/* used by rxe which owns the index */
@@ -205,9 +203,7 @@ static inline void queue_advance_producer(struct rxe_queue *q,
 		break;
 	case QUEUE_TYPE_TO_ULP:
 		/* used by ulp, rxe owns the index */
-		if (WARN_ON(1))
-			pr_warn("%s: attempt to advance driver index\n",
-				__func__);
+		WARN_ON(1);
 		break;
 	}
 }
@@ -227,15 +223,11 @@ static inline void queue_advance_consumer(struct rxe_queue *q,
 		break;
 	case QUEUE_TYPE_TO_CLIENT:
 		/* used by rxe, client owns the index */
-		if (WARN_ON(1))
-			pr_warn("%s: attempt to advance client index\n",
-				__func__);
+		WARN_ON(1);
 		break;
 	case QUEUE_TYPE_FROM_ULP:
 		/* used by ulp, rxe owns the index */
-		if (WARN_ON(1))
-			pr_warn("%s: attempt to advance driver index\n",
-				__func__);
+		WARN_ON(1);
 		break;
 	case QUEUE_TYPE_TO_ULP:
 		/* used by ulp which owns the index */
Signature · 1 line
2.27.0
maintainer
#1 lore raw
Quoted 10 lines · Zhu Yanjun

From: Zhu Yanjun <[email protected]>

Add the function names to the pr_ logs. As such, if some bugs occur, with function names, it is easy to locate the bugs.

---
 drivers/infiniband/sw/rxe/rxe.h       |  2 +-
 drivers/infiniband/sw/rxe/rxe_queue.h | 16 ++++------------
 2 files changed, 5 insertions(+), 13 deletions(-)

After you delete the warn_on's there is very little left:

rivers/infiniband/sw/rxe/rxe.c:                pr_err("rxe creation allowed on top of a real device only\n");
drivers/infiniband/sw/rxe/rxe.c:        pr_info("loaded\n");
drivers/infiniband/sw/rxe/rxe.c:        pr_info("unloaded\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to create IPv4 UDP tunnel\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_warn("IPv6 is not supported, can not create a UDPv6 socket\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to create IPv6 UDP tunnel\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to register netdev notifier\n");

It is not exactly mysterious where these come from?

Why do we need the function name?

Jason

#2 lore raw
Quoted 21 lines · Jason Gunthorpe
Quoted 10 lines · Zhu Yanjun

From: Zhu Yanjun <[email protected]>

Add the function names to the pr_ logs. As such, if some bugs occur, with function names, it is easy to locate the bugs.

---
  drivers/infiniband/sw/rxe/rxe.h       |  2 +-
  drivers/infiniband/sw/rxe/rxe_queue.h | 16 ++++------------
  2 files changed, 5 insertions(+), 13 deletions(-)

After you delete the warn_on's there is very little left:

rivers/infiniband/sw/rxe/rxe.c:                pr_err("rxe creation allowed on top of a real device only\n");
drivers/infiniband/sw/rxe/rxe.c:        pr_info("loaded\n");
drivers/infiniband/sw/rxe/rxe.c:        pr_info("unloaded\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to create IPv4 UDP tunnel\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_warn("IPv6 is not supported, can not create a UDPv6 socket\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to create IPv6 UDP tunnel\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to register netdev notifier\n");
Quoted 1 line

It is not exactly mysterious where these come from?

If I got you correctly, you mean that we can easily locate the above pr_xxx logs. As such, it is not necessary to add function names.

From this perspective, I agree with you.

Why I make this commit is that we can directly use this pr_info("xxxxx") when some bug occurs. But in the logs, the function names appear.

With this commit, we can decrease some work loads with some bug occurs.

Now I am debugging a problem. With a similar commit with this, I directly use pr_warn("xxx"). But in the logs, function names, line number, file names, caller and so on will appear.

This greatly decrease my workloads.

I think this is a smart method. So I introduce this smart method to softroce. When some bugs occur in softroce, this method can help us.

Zhu Yanjun

Quoted 4 lines

Why do we need the function name?

Jason

maintainer
#3 lore raw
Quoted 28 lines · Zhu Yanjun
Quoted 21 lines · Jason Gunthorpe
Quoted 10 lines · Zhu Yanjun

From: Zhu Yanjun <[email protected]>

Add the function names to the pr_ logs. As such, if some bugs occur, with function names, it is easy to locate the bugs.

---
  drivers/infiniband/sw/rxe/rxe.h       |  2 +-
  drivers/infiniband/sw/rxe/rxe_queue.h | 16 ++++------------
  2 files changed, 5 insertions(+), 13 deletions(-)

After you delete the warn_on's there is very little left:

rivers/infiniband/sw/rxe/rxe.c:                pr_err("rxe creation allowed on top of a real device only\n");
drivers/infiniband/sw/rxe/rxe.c:        pr_info("loaded\n");
drivers/infiniband/sw/rxe/rxe.c:        pr_info("unloaded\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to create IPv4 UDP tunnel\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_warn("IPv6 is not supported, can not create a UDPv6 socket\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to create IPv6 UDP tunnel\n");
drivers/infiniband/sw/rxe/rxe_net.c:            pr_err("Failed to register netdev notifier\n");
Quoted 1 line

It is not exactly mysterious where these come from?

If I got you correctly, you mean that we can easily locate the above pr_xxx logs. As such, it is not necessary to add function names.

Quoted 6 lines

From this perspective, I agree with you.

Why I make this commit is that we can directly use this pr_info("xxxxx") when some bug occurs. But in the logs, the function names appear.

With this commit, we can decrease some work loads with some bug occurs.

That doesn't seem like a good reason to merge something so abnormal for pr_* functions

Jason