Skip to contentSkip to conversation

Three fixes for the UPIU trace

PATCH In review Bean Huo 4 participants 20 messages Dec 6, 2020 – Dec 8, 2020

Subsystems (2)
lore
Share & tools
lore

Bean Huo

author
#0 lore raw

cover Three fixes for the UPIU trace

From: Bean Huo <[email protected]>

Bean Huo (3):
scsi: ufs: Distinguish between query REQ and query RSP in query trace
scsi: ufs: Distinguish between TM request UPIU and response UPIU in TM
UPIU trace
scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

 drivers/scsi/ufs/ufshcd.c  | 21 ++++++++++++++++-----
 include/trace/events/ufs.h | 10 +++++++---
 2 files changed, 23 insertions(+), 8 deletions(-)
Signature · 1 line
2.17.1
author
#1 lore raw

PATCH 1/3 scsi: ufs: Distinguish between query REQ and query RSP in query trace

From: Bean Huo <[email protected]>

Currently, in the query completion trace print, since we use hba->lrb[tag].ucd_req_ptr and didn't differentiate UPIU between request and response, thus header and transaction-specific field in UPIU printed by query trace are identical. This is not very practical. As below:

query_send: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00
query_complete: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00

For the failure analysis, we want to understand the real response reported by the UFS device, however, the current query trace tells us nothing. After this patch, the query trace on the query_send, and the above a pair of query_send and query_complete will be:

query_send: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00
ufshcd_upiu: HDR:36 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 01 00 00 00 00

---
 drivers/scsi/ufs/ufshcd.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/scsi/ufs/ufshcd.c +8 -2
index ceb562accc39..e10de94adb3f 100644
@@ -321,9 +321,15 @@ static void ufshcd_add_cmd_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 		const char *str)
 {
-	struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
+	struct utp_upiu_req *rq_rsp;
+
+	if (!strcmp("query_send", str))
+		rq_rsp = hba->lrb[tag].ucd_req_ptr;
+	else
+		rq_rsp = (struct utp_upiu_req *)hba->lrb[tag].ucd_rsp_ptr;
 
-	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq->qr);
+	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq_rsp->header,
+			  &rq_rsp->qr);
 }
 
 static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag,
Signature · 1 line
2.17.1
#2 lore raw

Re: [PATCH v1 1/3] scsi: ufs: Distinguish between query REQ and query RSP in query trace

Quoted 34 lines · Bean Huo

From: Bean Huo <[email protected]>

Currently, in the query completion trace print, since we use hba->lrb[tag].ucd_req_ptr and didn't differentiate UPIU between request and response, thus header and transaction-specific field in UPIU printed by query trace are identical. This is not very practical. As below:

query_send: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00
query_complete: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00

For the failure analysis, we want to understand the real response reported by the UFS device, however, the current query trace tells us nothing. After this patch, the query trace on the query_send, and the above a pair of query_send and query_complete will be:

query_send: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00
ufshcd_upiu: HDR:36 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 01 00 00 00 00

---
 drivers/scsi/ufs/ufshcd.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index ceb562accc39..e10de94adb3f 100644
@@ -321,9 +321,15 @@ static void ufshcd_add_cmd_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 		const char *str)
 {
-	struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
+	struct utp_upiu_req *rq_rsp;

drivers/scsi/ufs/ufshcd.c static void ufshcd_add_cmd_upiu_trace(struct ufs_hba *hba, unsigned int tag, · line 333

Quoted 10 lines
drivers/scsi/ufs/ufshcd.c +7 -1
+
+	if (!strcmp("query_send", str))
+		rq_rsp = hba->lrb[tag].ucd_req_ptr;
+	else
+		rq_rsp = (struct utp_upiu_req *)hba->lrb[tag].ucd_rsp_ptr;
 
-	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq->qr);
+	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq_rsp->header,
+			  &rq_rsp->qr);
 }

Please change the 'str' argument into an enum and let ufshcd_add_query_upiu_trace() do the enum-to-string conversion. That will allow to convert the strcmp() call into an integer comparison.

Thanks,

Bart.

Acked-by on v1 1/3
#3 lore raw

RE: [PATCH v1 1/3] scsi: ufs: Distinguish between query REQ and query RSP in query trace

In reply to Bean Huo
Quoted 19 lines

From: Bean Huo <[email protected]>

Currently, in the query completion trace print, since we use hba->lrb[tag].ucd_req_ptr and didn't differentiate UPIU between request and response, thus header and transaction-specific field in UPIU printed by query trace are identical. This is not very practical. As below:

query_send: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00
query_complete: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00

For the failure analysis, we want to understand the real response reported by the UFS device, however, the current query trace tells us nothing. After this patch, the query trace on the query_send, and the above a pair of query_send and query_complete will be:

Quoted 6 lines

query_send: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00
ufshcd_upiu: HDR:36 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 01 00 00 00 00

But you need to change the complete string so not to break the current parsers. I would also pass to ufshcd_add_query_upiu_trace the struct utp_upiu_req *, so no comparison is needed.

Thanks,
Avri

Quoted 31 lines
---
 drivers/scsi/ufs/ufshcd.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/scsi/ufs/ufshcd.c +8 -2
index ceb562accc39..e10de94adb3f 100644
@@ -321,9 +321,15 @@ static void ufshcd_add_cmd_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
 static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int
 tag,
                const char *str)
 {
-       struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
+       struct utp_upiu_req *rq_rsp;
+
+       if (!strcmp("query_send", str))
+               rq_rsp = hba->lrb[tag].ucd_req_ptr;
+       else
+               rq_rsp = (struct utp_upiu_req *)hba->lrb[tag].ucd_rsp_ptr;
 
-       trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq->qr);
+       trace_ufshcd_upiu(dev_name(hba->dev), str, &rq_rsp->header,
+                         &rq_rsp->qr);
 }

static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int
tag,
--
2.17.1

maintainer
#4 lore raw

Re: [PATCH v1 1/3] scsi: ufs: Distinguish between query REQ and query RSP in query trace

In reply to Bean Huo
Quoted 29 lines · On Sun, 6 Dec 2020 17:42:24 +0100 Bean Huo <[email protected]> wrote:

From: Bean Huo <[email protected]>

Currently, in the query completion trace print, since we use hba->lrb[tag].ucd_req_ptr and didn't differentiate UPIU between request and response, thus header and transaction-specific field in UPIU printed by query trace are identical. This is not very practical. As below:

query_send: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00
query_complete: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00

For the failure analysis, we want to understand the real response reported by the UFS device, however, the current query trace tells us nothing. After this patch, the query trace on the query_send, and the above a pair of query_send and query_complete will be:

query_send: HDR:16 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 00 00 00 00 00
ufshcd_upiu: HDR:36 00 00 0e 00 81 00 00 00 00 00 00, CDB:06 0e 03 00 00 00 00 00 00 00 00 01 00 00 00 00

---
 drivers/scsi/ufs/ufshcd.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/scsi/ufs/ufshcd.c +0 -0
index ceb562accc39..e10de94adb3f 100644

drivers/scsi/ufs/ufshcd.c static void ufshcd_add_cmd_upiu_trace(struct ufs_hba *hba, unsigned int tag, · line 325

Quoted 6 lines
drivers/scsi/ufs/ufshcd.c +2 -1
@@ -321,9 +321,15 @@ static void ufshcd_add_cmd_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 		const char *str)
 {
-	struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
+	struct utp_upiu_req *rq_rsp;
+

I would add:

	if (!trace_ufshcd_upiu_enabled())
		return;

Why do the work if the trace point is not enabled?

-- Steve

Quoted 11 lines
drivers/scsi/ufs/ufshcd.c +6 -1
+	if (!strcmp("query_send", str))
+		rq_rsp = hba->lrb[tag].ucd_req_ptr;
+	else
+		rq_rsp = (struct utp_upiu_req *)hba->lrb[tag].ucd_rsp_ptr;
 
-	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq->qr);
+	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq_rsp->header,
+			  &rq_rsp->qr);
 }
 
 static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag,
author
#5 lore raw

Re: [PATCH v1 1/3] scsi: ufs: Distinguish between query REQ and query RSP in query trace

Quoted 12 lines · Steven Rostedt

drivers/scsi/ufs/ufshcd.c static void ufshcd_add_cmd_upiu_trace(struct · line 327

Quoted 9 lines
drivers/scsi/ufs/ufshcd.c +2 -1
@@ -321,9 +321,15 @@ static void ufshcd_add_cmd_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
  static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba,
 unsigned int tag,
               const char *str)
  {
-     struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
+     struct utp_upiu_req *rq_rsp;
+

I would add:

Quoted 6 lines
        if (!trace_ufshcd_upiu_enabled())
                return;

Why do the work if the trace point is not enabled?

-- Steve

Steve,

Thanks a lot, I will fix it in the next version.

Thanks,
Bean

author
#6 lore raw

PATCH 2/3 scsi: ufs: Distinguish between TM request UPIU and response UPIU in TM UPIU trace

In reply to Bean Huo

From: Bean Huo <[email protected]>

Distinguish between TM request UPIU and response UPIU in TM UPIU trace, for the TM response, let TM UPIU trace print its TM response UPIU.

---
 drivers/scsi/ufs/ufshcd.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/scsi/ufs/ufshcd.c +6 -2
index e10de94adb3f..29d7240a61bf 100644
@@ -338,8 +338,12 @@ static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 	int off = (int)tag - hba->nutrs;
 	struct utp_task_req_desc *descp = &hba->utmrdl_base_addr[off];
 
-	trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->req_header,
-			&descp->input_param1);
+	if (!strcmp("tm_send", str))
+		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->req_header,
+				  &descp->input_param1);
+	else
+		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->rsp_header,
+				  &descp->output_param1);
 }
 
 static void ufshcd_add_uic_command_trace(struct ufs_hba *hba,
Signature · 1 line
2.17.1
#7 lore raw

Re: [PATCH v1 2/3] scsi: ufs: Distinguish between TM request UPIU and response UPIU in TM UPIU trace

Quoted 17 lines · Bean Huo

From: Bean Huo <[email protected]>

Distinguish between TM request UPIU and response UPIU in TM UPIU trace, for the TM response, let TM UPIU trace print its TM response UPIU.

---
 drivers/scsi/ufs/ufshcd.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/scsi/ufs/ufshcd.c +0 -0
index e10de94adb3f..29d7240a61bf 100644
@@ -338,8 +338,12 @@ static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 	int off = (int)tag - hba->nutrs;
 	struct utp_task_req_desc *descp = &hba->utmrdl_base_addr[off];

drivers/scsi/ufs/ufshcd.c static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag, · line 347

Quoted 10 lines
drivers/scsi/ufs/ufshcd.c +6 -2
 
-	trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->req_header,
-			&descp->input_param1);
+	if (!strcmp("tm_send", str))
+		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->req_header,
+				  &descp->input_param1);
+	else
+		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->rsp_header,
+				  &descp->output_param1);
 }

Same comment here: please change the type of the 'str' argument in an enum such that the strcmp() call can be changed into an integer comparison.

Thanks,

Bart.

Acked-by on v1 2/3
#8 lore raw

RE: [PATCH v1 2/3] scsi: ufs: Distinguish between TM request UPIU and response UPIU in TM UPIU trace

In reply to Bean Huo
Quoted 8 lines

From: Bean Huo <[email protected]>

Distinguish between TM request UPIU and response UPIU in TM UPIU trace, for the TM response, let TM UPIU trace print its TM response UPIU.

Again - same comment:
But you need to change the complete string so not to break the current parsers. I would also pass to the struct utp_upiu_header *,
so no comparison is needed.

Thanks,
Avri

Quoted 28 lines
---
 drivers/scsi/ufs/ufshcd.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/scsi/ufs/ufshcd.c +6 -2
index e10de94adb3f..29d7240a61bf 100644
@@ -338,8 +338,12 @@ static void ufshcd_add_tm_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
        int off = (int)tag - hba->nutrs;
        struct utp_task_req_desc *descp = &hba->utmrdl_base_addr[off];
 
-       trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->req_header,
-                       &descp->input_param1);
+       if (!strcmp("tm_send", str))
+               trace_ufshcd_upiu(dev_name(hba->dev), str, &descp-
 >req_header,
+                                 &descp->input_param1);
+       else
+               trace_ufshcd_upiu(dev_name(hba->dev), str, &descp-
 >rsp_header,
+                                 &descp->output_param1);

}

static void ufshcd_add_uic_command_trace(struct ufs_hba *hba,
--
2.17.1

author
#9 lore raw

PATCH 3/3 scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

In reply to Bean Huo

From: Bean Huo <[email protected]>

Transaction Specific Fields (TSF) in the UPIU package could be CDB (SCSI/UFS Command Descriptor Block), OSF (Opcode Specific Field), and TM I/O parameter (Task Management Input/Output Parameter). But, currently, we take all of these as CDB in the UPIU trace. Thus makes user confuse among CDB, OSF, and TM message. So fix it with this patch.

---
 drivers/scsi/ufs/ufshcd.c  |  9 +++++----
 include/trace/events/ufs.h | 10 +++++++---
 2 files changed, 12 insertions(+), 7 deletions(-)
drivers/scsi/ufs/ufshcd.c +5 -4
index 29d7240a61bf..5b2219e44743 100644
@@ -315,7 +315,8 @@ static void ufshcd_add_cmd_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 {
 	struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
 
-	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq->sc.cdb);
+	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq->sc.cdb,
+			  "CDB");
 }
 
 static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int tag,
@@ -329,7 +330,7 @@ static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 		rq_rsp = (struct utp_upiu_req *)hba->lrb[tag].ucd_rsp_ptr;
 
 	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq_rsp->header,
-			  &rq_rsp->qr);
+			  &rq_rsp->qr, "OSF");
 }
 
 static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag,
@@ -340,10 +341,10 @@ static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 
 	if (!strcmp("tm_send", str))
 		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->req_header,
-				  &descp->input_param1);
+				  &descp->input_param1, "TM_INPUT");
 	else
 		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->rsp_header,
-				  &descp->output_param1);
+				  &descp->output_param1, "TM_OUTPUT");
 }
 
 static void ufshcd_add_uic_command_trace(struct ufs_hba *hba,
include/trace/events/ufs.h +7 -3
index 0bd54a184391..68e8e97a9b47 100644
@@ -295,15 +295,17 @@ TRACE_EVENT(ufshcd_uic_command,
 );
 
 TRACE_EVENT(ufshcd_upiu,
-	TP_PROTO(const char *dev_name, const char *str, void *hdr, void *tsf),
+	TP_PROTO(const char *dev_name, const char *str, void *hdr, void *tsf,
+		 const char *tsf_type),
 
-	TP_ARGS(dev_name, str, hdr, tsf),
+	TP_ARGS(dev_name, str, hdr, tsf, tsf_type),
 
 	TP_STRUCT__entry(
 		__string(dev_name, dev_name)
 		__string(str, str)
 		__array(unsigned char, hdr, 12)
 		__array(unsigned char, tsf, 16)
+		__string(tsf_type, tsf_type)
 	),
 
 	TP_fast_assign(
@@ -311,12 +313,14 @@ TRACE_EVENT(ufshcd_upiu,
 		__assign_str(str, str);
 		memcpy(__entry->hdr, hdr, sizeof(__entry->hdr));
 		memcpy(__entry->tsf, tsf, sizeof(__entry->tsf));
+		__assign_str(tsf_type, tsf_type);
 	),
 
 	TP_printk(
-		"%s: %s: HDR:%s, CDB:%s",
+		"%s: %s: HDR:%s, %s:%s",
 		__get_str(str), __get_str(dev_name),
 		__print_hex(__entry->hdr, sizeof(__entry->hdr)),
+		__get_str(tsf_type),
 		__print_hex(__entry->tsf, sizeof(__entry->tsf))
 	)
 );
Signature · 1 line
2.17.1
#10 lore raw

RE: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

Quoted 87 lines

From: Bean Huo <[email protected]>

Transaction Specific Fields (TSF) in the UPIU package could be CDB (SCSI/UFS Command Descriptor Block), OSF (Opcode Specific Field), and TM I/O parameter (Task Management Input/Output Parameter). But, currently,
we take all of these as CDB in the UPIU trace. Thus makes user confuse among CDB, OSF, and TM message. So fix it with this patch.

---
 drivers/scsi/ufs/ufshcd.c  |  9 +++++----
 include/trace/events/ufs.h | 10 +++++++---
 2 files changed, 12 insertions(+), 7 deletions(-)
drivers/scsi/ufs/ufshcd.c +2 -1
index 29d7240a61bf..5b2219e44743 100644
@@ -315,7 +315,8 @@ static void ufshcd_add_cmd_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
 {
        struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
 
-       trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq-
 >sc.cdb);
+       trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq-
 >sc.cdb,
+                         "CDB");

}

static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int
tag,

drivers/scsi/ufs/ufshcd.c +1 -1
@@ -329,7 +330,7 @@ static void ufshcd_add_query_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
                rq_rsp = (struct utp_upiu_req *)hba->lrb[tag].ucd_rsp_ptr;
 
        trace_ufshcd_upiu(dev_name(hba->dev), str, &rq_rsp->header,
-                         &rq_rsp->qr);
+                         &rq_rsp->qr, "OSF");
 }
 

static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int
tag,

drivers/scsi/ufs/ufshcd.c +2 -2
@@ -340,10 +341,10 @@ static void ufshcd_add_tm_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
 
        if (!strcmp("tm_send", str))
                trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->req_header,
-                                 &descp->input_param1);
+                                 &descp->input_param1, "TM_INPUT");
        else
                trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->rsp_header,
-                                 &descp->output_param1);
+                                 &descp->output_param1, "TM_OUTPUT");
 }
 

static void ufshcd_add_uic_command_trace(struct ufs_hba *hba,

include/trace/events/ufs.h +4 -2
index 0bd54a184391..68e8e97a9b47 100644
@@ -295,15 +295,17 @@ TRACE_EVENT(ufshcd_uic_command,
 );
 
 TRACE_EVENT(ufshcd_upiu,
-       TP_PROTO(const char *dev_name, const char *str, void *hdr, void *tsf),
+       TP_PROTO(const char *dev_name, const char *str, void *hdr, void *tsf,
+                const char *tsf_type),
 
-       TP_ARGS(dev_name, str, hdr, tsf),
+       TP_ARGS(dev_name, str, hdr, tsf, tsf_type),
 
        TP_STRUCT__entry(
                __string(dev_name, dev_name)
                __string(str, str)
                __array(unsigned char, hdr, 12)
                __array(unsigned char, tsf, 16)
+               __string(tsf_type, tsf_type)
        ),
 
        TP_fast_assign(
@@ -311,12 +313,14 @@ TRACE_EVENT(ufshcd_upiu,
                __assign_str(str, str);
                memcpy(__entry->hdr, hdr, sizeof(__entry->hdr));

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 323

Quoted 10 lines
include/trace/events/ufs.h +3 -1
                memcpy(__entry->tsf, tsf, sizeof(__entry->tsf));
+               __assign_str(tsf_type, tsf_type);
        ),
 
        TP_printk(
-               "%s: %s: HDR:%s, CDB:%s",
+               "%s: %s: HDR:%s, %s:%s",
                __get_str(str), __get_str(dev_name),
                __print_hex(__entry->hdr, sizeof(__entry->hdr)),
+               __get_str(tsf_type),

This breaks what current parsers expects.
Why str is not enough to distinguish between the command?

Thanks,
Avri

Quoted 5 lines
include/trace/events/ufs.h +0 -1
                __print_hex(__entry->tsf, sizeof(__entry->tsf))
        )
 );
--

2.17.1

author
#11 lore raw

Re: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

Quoted 7 lines · Avri Altman

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 323

Quoted 6 lines
         TP_printk(
include/trace/events/ufs.h +2 -1
-               "%s: %s: HDR:%s, CDB:%s",
+               "%s: %s: HDR:%s, %s:%s",
                 __get_str(str), __get_str(dev_name),
                 __print_hex(__entry->hdr, sizeof(__entry->hdr)),
+               __get_str(tsf_type),
Quoted 5 lines

This breaks what current parsers expects.
Why str is not enough to distinguish between the command?

Thanks,
Avri

Hi Avri
Tt donesn't break original CDB parser. for the CDB, it is still the same as before. Here just make Transaction Specific Fields in the UPIU package much clearer.

I mentioned in the commits message:

Transaction Specific Fields (TSF) in the UPIU package could be CDB (SCSI/UFS Command Descriptor Block), OSF (Opcode Specific Field), and TM I/O parameter (Task Management Input/Output Parameter). But we didn't differenciate them. we take all of these as CDB. This is wrong.

I want to make it clearer and make UPIU trace in line with the Spec. what's more, how do you filter OSF, TM parameters with current UPIU trace? you take all of them as CDB? if so, I think, it's better to change parser.

Thanks,
Bean

#12 lore raw

RE: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

Quoted 12 lines · Avri Altman
Quoted 10 lines

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 323

Quoted 6 lines
         TP_printk(
include/trace/events/ufs.h +2 -1
-               "%s: %s: HDR:%s, CDB:%s",
+               "%s: %s: HDR:%s, %s:%s",
                 __get_str(str), __get_str(dev_name),
                 __print_hex(__entry->hdr, sizeof(__entry->hdr)),
+               __get_str(tsf_type),

This breaks what current parsers expects.
Why str is not enough to distinguish between the command?

Quoted 7 lines
Quoted 2 lines

Thanks,
Avri

Hi Avri
Tt donesn't break original CDB parser. for the CDB, it is still the same as before. Here just make Transaction Specific Fields in the UPIU package much clearer.

It does breaks our current parser that expects "CDB:" for all upiu types

Quoted 8 lines

I mentioned in the commits message:

Transaction Specific Fields (TSF) in the UPIU package could be CDB (SCSI/UFS Command Descriptor Block), OSF (Opcode Specific Field), and TM I/O parameter (Task Management Input/Output Parameter). But we didn't differenciate them. we take all of these as CDB. This is wrong.

Quoted 4 lines

I want to make it clearer and make UPIU trace in line with the Spec. what's more, how do you filter OSF, TM parameters with current UPIU trace? you take all of them as CDB? if so, I think, it's better to change parser.

Indeed, it is just a small change, but breaking user-space is not an acceptable approach. Also, the upiu tracer was never meant to be human-readable: it just dump the upiu, Which contains all the info required to parse it anyway,
So breaking user-space just to making it more readable doesn't really make sense?

Looking at the previous 2 patches of this series, I was hoping that you will do the same for Command upiu, as well?
Again - same comment: if you are doing that need to change the str not to break current parsers.

Thanks,
Avri

Quoted 8 lines

Thanks,
Bean

author
#13 lore raw

Re: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

Quoted 18 lines · Avri Altman
Quoted 9 lines

didn't differenciate them. we take all of these as CDB. This is wrong.

I want to make it clearer and make UPIU trace in line with the Spec.
what's more, how do you filter OSF, TM parameters with current UPIU
trace? you take all of them as CDB? if so, I think, it's better to change parser.

Indeed, it is just a small change, but breaking user-space is not an acceptable approach.
Also, the upiu tracer was never meant to be human-readable: it just dump the upiu,
Which contains all the info required to parse it anyway, So breaking user-space just to making it more readable doesn't really make sense?

Quoted 8 lines

Looking at the previous 2 patches of this series, I was hoping that you will do the same for
Command upiu, as well?
Again - same comment: if you are doing that need to change the str not to break current parsers.

Thanks,
Avri

will not change original CDB format, just add new OSF, TM. the string format will not be change. The current the HDR and CDB in the send and complete trace are the same, I guess, you even didn't trace CDB in your parser, they cannot tell you the request execution result.

Bean

maintainer
#14 lore raw

Re: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

In reply to Avri Altman
Quoted 9 lines · On Mon, 7 Dec 2020 07:57:27 +0000 Avri Altman <[email protected]> wrote:

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 323

Quoted 7 lines
        TP_printk(
include/trace/events/ufs.h +2 -1
-               "%s: %s: HDR:%s, CDB:%s",
+               "%s: %s: HDR:%s, %s:%s",
                __get_str(str), __get_str(dev_name),
                __print_hex(__entry->hdr, sizeof(__entry->hdr)),
+               __get_str(tsf_type),  

This breaks what current parsers expects.
Why str is not enough to distinguish between the command?

Hopefully it shouldn't. Reading from user space should use the libtraceevent library, that reads the format files and extracts the raw data to find the fields. As long as the field exists, it should not break user space parsers. If it does, please let me know, and I'll gladly help change the user space code to use libtraceevent :-)

-- Steve

#15 lore raw

RE: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

Quoted 11 lines
Quoted 7 lines · On Mon, 7 Dec 2020 07:57:27 +0000 Avri Altman <[email protected]> wrote:
Quoted 7 lines
        TP_printk(
include/trace/events/ufs.h +2 -1
-               "%s: %s: HDR:%s, CDB:%s",
+               "%s: %s: HDR:%s, %s:%s",
                __get_str(str), __get_str(dev_name),
                __print_hex(__entry->hdr, sizeof(__entry->hdr)),
+               __get_str(tsf_type),
Quoted 8 lines
Quoted 2 lines

This breaks what current parsers expects.
Why str is not enough to distinguish between the command?

Hopefully it shouldn't. Reading from user space should use the libtraceevent library, that reads the format files and extracts the raw data to find the fields. As long as the field exists, it should not break user space parsers. If it does, please let me know, and I'll gladly help change the user space code to use libtraceevent :-)

Hi Steve,
Thanks. I wasn't aware of libtraceevent - is this a new thing?

We have a relatively sophisticated analysis platform that utilizes raw traces, Among which the upiu trace is the most important and informative.

This tool has evolved over the years, adding more and more parsers per need, and the users are picking the appropriate parser per the trace they used.

We will surely be glad to adopt new tracing capabilities, But we would prefer not to break anything.

Thanks,
Avri

maintainer
#16 lore raw

Re: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

Quoted 14 lines · On Mon, 7 Dec 2020 16:40:51 +0000 Avri Altman <[email protected]> wrote:
Quoted 14 lines
Quoted 9 lines · On Mon, 7 Dec 2020 07:57:27 +0000 Avri Altman <[email protected]> wrote:

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 323

Quoted 7 lines
        TP_printk(
include/trace/events/ufs.h +2 -1
-               "%s: %s: HDR:%s, CDB:%s",
+               "%s: %s: HDR:%s, %s:%s",
                __get_str(str), __get_str(dev_name),
                __print_hex(__entry->hdr, sizeof(__entry->hdr)),
+               __get_str(tsf_type),  

This breaks what current parsers expects.
Why str is not enough to distinguish between the command?

Quoted 7 lines
Quoted 5 lines

Hopefully it shouldn't. Reading from user space should use the libtraceevent library, that reads the format files and extracts the raw data to find the fields. As long as the field exists, it should not break user space parsers. If it does, please let me know, and I'll gladly help change the user space code to use libtraceevent :-)

Hi Steve,
Thanks. I wasn't aware of libtraceevent - is this a new thing?

Actually, it's been around almost as long as ftrace. But unfortunately, it's just now becoming a separate library. It was originally developed for trace-cmd, but has been copied into perf, power-top and rasdaemon. But this copying is inefficient and a maintenance nightmare, and we finally have the library as a stand alone, and hopefully will be delivered by distributions (I believe they are packaging it).

https://git.kernel.org/pub/scm/libs/libtrace/libtraceevent.git/

Looks like distros are starting to catch on.

https://packages.debian.org/unstable/libtraceevent-dev

We are currently working on libtracefs

https://git.kernel.org/pub/scm/libs/libtrace/libtracefs.git/

Which will make it a lot easier for applications to interact with the tracefs file system. I'm hoping to have this ready for distros by the end of the year. We have applications coming that depend on these.

Quoted 8 lines

We have a relatively sophisticated analysis platform that utilizes raw traces, Among which the upiu trace is the most important and informative.

This tool has evolved over the years, adding more and more parsers per need, and the users are picking the appropriate parser per the trace they used.

We will surely be glad to adopt new tracing capabilities,

I think libtraceevent and libtracefs would be a much welcome addition for upiu trace as it would be reading raw data (very fast), and have an API that makes doing so much simpler. For example, I just wrote a quick program that checks what files an application opens (this is not in anyway production ready):

http://rostedt.org/code/show-open-files.c

Quoted 1 line

But we would prefer not to break anything.

Of course!

And again, I would be happy to help out in converting to this libraries. It will make your applications more robust, as they make it so that you do not need to rely on the order of fields.

Note, there's plans on making these libraries python modules as well (to have python scripts enable and read ftrace data).

-- Steve

#17 lore raw

RE: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

Quoted 69 lines
Quoted 14 lines · On Mon, 7 Dec 2020 16:40:51 +0000 Avri Altman <[email protected]> wrote:
Quoted 14 lines
Quoted 9 lines · On Mon, 7 Dec 2020 07:57:27 +0000 Avri Altman <[email protected]> wrote:

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 323

Quoted 7 lines
        TP_printk(
include/trace/events/ufs.h +2 -1
-               "%s: %s: HDR:%s, CDB:%s",
+               "%s: %s: HDR:%s, %s:%s",
                __get_str(str), __get_str(dev_name),
                __print_hex(__entry->hdr, sizeof(__entry->hdr)),
+               __get_str(tsf_type),

This breaks what current parsers expects.
Why str is not enough to distinguish between the command?

Quoted 7 lines
Quoted 5 lines

Hopefully it shouldn't. Reading from user space should use the libtraceevent library, that reads the format files and extracts the raw data to find the fields. As long as the field exists, it should not break user space parsers. If it does, please let me know, and I'll gladly help change the user space code to use libtraceevent :-)

Hi Steve,
Thanks. I wasn't aware of libtraceevent - is this a new thing?

Actually, it's been around almost as long as ftrace. But unfortunately, it's just now becoming a separate library. It was originally developed for trace-cmd, but has been copied into perf, power-top and rasdaemon. But this
copying is inefficient and a maintenance nightmare, and we finally have the library as a stand alone, and hopefully will be delivered by distributions (I believe they are packaging it).

https://git.kernel.org/pub/scm/libs/libtrace/libtraceevent.git/

Looks like distros are starting to catch on.

https://packages.debian.org/unstable/libtraceevent-dev

We are currently working on libtracefs

https://git.kernel.org/pub/scm/libs/libtrace/libtracefs.git/

Which will make it a lot easier for applications to interact with the tracefs file system. I'm hoping to have this ready for distros by the end of the year. We have applications coming that depend on these.

Quoted 5 lines

We have a relatively sophisticated analysis platform that utilizes raw traces, Among which the upiu trace is the most important and informative.

This tool has evolved over the years, adding more and more parsers per

need,

Quoted 3 lines

and the users are picking the appropriate parser per the trace they used.

We will surely be glad to adopt new tracing capabilities,

I think libtraceevent and libtracefs would be a much welcome addition for upiu trace as it would be reading raw data (very fast), and have an API that makes doing so much simpler. For example, I just wrote a quick program that checks what files an application opens (this is not in anyway production ready):

http://rostedt.org/code/show-open-files.c

Quoted 1 line

But we would prefer not to break anything.

Quoted 8 lines

Of course!

And again, I would be happy to help out in converting to this libraries. It will make your applications more robust, as they make it so that you do not need to rely on the order of fields.

Note, there's plans on making these libraries python modules as well (to have python scripts enable and read ftrace data).

Thanks a lot for your insightful comments.
We will surely look into libtraceevent and libtracefs

Thanks,
Avri

Quoted 2 lines

-- Steve

maintainer
#18 lore raw

Re: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

In reply to Bean Huo
Quoted 38 lines · On Sun, 6 Dec 2020 17:42:26 +0100 Bean Huo <[email protected]> wrote:

From: Bean Huo <[email protected]>

Transaction Specific Fields (TSF) in the UPIU package could be CDB (SCSI/UFS Command Descriptor Block), OSF (Opcode Specific Field), and TM I/O parameter (Task Management Input/Output Parameter). But, currently, we take all of these as CDB in the UPIU trace. Thus makes user confuse among CDB, OSF, and TM message. So fix it with this patch.

---
 drivers/scsi/ufs/ufshcd.c  |  9 +++++----
 include/trace/events/ufs.h | 10 +++++++---
 2 files changed, 12 insertions(+), 7 deletions(-)
drivers/scsi/ufs/ufshcd.c +3 -2
index 29d7240a61bf..5b2219e44743 100644
@@ -315,7 +315,8 @@ static void ufshcd_add_cmd_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 {
 	struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
 
-	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq->sc.cdb);
+	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq->sc.cdb,
+			  "CDB");
 }
 
 static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int tag,
@@ -329,7 +330,7 @@ static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 		rq_rsp = (struct utp_upiu_req *)hba->lrb[tag].ucd_rsp_ptr;
 
 	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq_rsp->header,
-			  &rq_rsp->qr);
+			  &rq_rsp->qr, "OSF");
 }
 
 static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag,

drivers/scsi/ufs/ufshcd.c static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag, · line 347

Quoted 9 lines
drivers/scsi/ufs/ufshcd.c +2 -2
@@ -340,10 +341,10 @@ static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned int tag,
 
 	if (!strcmp("tm_send", str))
 		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->req_header,
-				  &descp->input_param1);
+				  &descp->input_param1, "TM_INPUT");
 	else
 		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp->rsp_header,
-				  &descp->output_param1);
+				  &descp->output_param1, "TM_OUTPUT");

You could save some space on the ring buffer, if you made the above into an enum, and then used print_symbolic().

include/trace/events/ufs.h TRACE_EVENT(ufshcd_uic_command, · line 295

Quoted 9 lines
drivers/scsi/ufs/ufshcd.c +0 -0
 }
 
 static void ufshcd_add_uic_command_trace(struct ufs_hba *hba,
include/trace/events/ufs.h +0 -0
index 0bd54a184391..68e8e97a9b47 100644
@@ -295,15 +295,17 @@ TRACE_EVENT(ufshcd_uic_command,
 );

You could make this:

#define TRACE_TSF_TYPES		\
	EM(CDB)			\
	EM(OSF)			\
	EM(TM_INPUT)		\
	EMe(TM_OUTPUT)

#ifndef TRACE_TSF_TYPES_ENUMS
#define TRACE_TSF_TYPES_ENUMS
#undef EM
#undef EMe

#define EM(x) TRACE_TSF_##x,
#define EMe(x) TRACE_TSF_##x

enum {
TRACE_TSF_TYPES
}
#endif /* TRACE_TSF_TYPES_ENUMS */

#undef EM
#undef EMe

/* These export the enum names to user space */
#define EM(x) TRACE_DEFINE_ENUM(TRACE_TSF_##x)
#define EMe(x) TRACE_DEFINE_ENUM(TRACE_TSF_##x)

TRACE_TSF_TYPES

#undef EM
#undef EMe

/* These are used in the print_symbolic */
#define EM(x) { TRACE_TSF_##x, #x },
#define EMe(x) { TRACE_TSF_##x, #x }

include/trace/events/ufs.h TRACE_EVENT(ufshcd_uic_command, · line 299

Quoted 5 lines
include/trace/events/ufs.h +2 -1
 TRACE_EVENT(ufshcd_upiu,
-	TP_PROTO(const char *dev_name, const char *str, void *hdr, void *tsf),
+	TP_PROTO(const char *dev_name, const char *str, void *hdr, void *tsf,
+		 const char *tsf_type),
		int tsf_type;

include/trace/events/ufs.h TRACE_EVENT(ufshcd_uic_command, · line 308

Quoted 10 lines
include/trace/events/ufs.h +2 -1
-	TP_ARGS(dev_name, str, hdr, tsf),
+	TP_ARGS(dev_name, str, hdr, tsf, tsf_type),
 
 	TP_STRUCT__entry(
 		__string(dev_name, dev_name)
 		__string(str, str)
 		__array(unsigned char, hdr, 12)
 		__array(unsigned char, tsf, 16)
+		__string(tsf_type, tsf_type)
		__field(int, tsf_type)

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 316

Quoted 8 lines
include/trace/events/ufs.h +1 -0
 	),
 
 	TP_fast_assign(
@@ -311,12 +313,14 @@ TRACE_EVENT(ufshcd_upiu,
 		__assign_str(str, str);
 		memcpy(__entry->hdr, hdr, sizeof(__entry->hdr));
 		memcpy(__entry->tsf, tsf, sizeof(__entry->tsf));
+		__assign_str(tsf_type, tsf_type);
		__entry->tsf_type = tsf_type;

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 323

Quoted 8 lines
include/trace/events/ufs.h +2 -1
 	),
 
 	TP_printk(
-		"%s: %s: HDR:%s, CDB:%s",
+		"%s: %s: HDR:%s, %s:%s",
 		__get_str(str), __get_str(dev_name),
 		__print_hex(__entry->hdr, sizeof(__entry->hdr)),
+		__get_str(tsf_type),

print_symbolic(tsf_type, TRACE_TSF_TYPES),

-- Steve

Quoted 3 lines
include/trace/events/ufs.h +0 -0
 		__print_hex(__entry->tsf, sizeof(__entry->tsf))
 	)
 );
author
#19 lore raw

Re: [PATCH v1 3/3] scsi: ufs: Make UPIU trace easier differentiate among CDB, OSF, and TM

Quoted 165 lines · Steven Rostedt
Quoted 50 lines · On Sun, 6 Dec 2020 17:42:26 +0100 Bean Huo <[email protected]> wrote:

From: Bean Huo <[email protected]>

Transaction Specific Fields (TSF) in the UPIU package could be CDB (SCSI/UFS Command Descriptor Block), OSF (Opcode Specific Field), and
TM I/O parameter (Task Management Input/Output Parameter). But, currently,
we take all of these as CDB in the UPIU trace. Thus makes user confuse
among CDB, OSF, and TM message. So fix it with this patch.

---
 drivers/scsi/ufs/ufshcd.c  |  9 +++++----
 include/trace/events/ufs.h | 10 +++++++---
 2 files changed, 12 insertions(+), 7 deletions(-)
drivers/scsi/ufs/ufshcd.c +2 -1
index 29d7240a61bf..5b2219e44743 100644
@@ -315,7 +315,8 @@ static void ufshcd_add_cmd_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
 {
 	struct utp_upiu_req *rq = hba->lrb[tag].ucd_req_ptr;
 
-	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq-
 >sc.cdb);
+	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq->header, &rq-
 >sc.cdb,
+			  "CDB");

}

static void ufshcd_add_query_upiu_trace(struct ufs_hba *hba,
unsigned int tag,

drivers/scsi/ufs/ufshcd.c +1 -1
@@ -329,7 +330,7 @@ static void ufshcd_add_query_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
 		rq_rsp = (struct utp_upiu_req *)hba-
 >lrb[tag].ucd_rsp_ptr;
 
 	trace_ufshcd_upiu(dev_name(hba->dev), str, &rq_rsp->header,
-			  &rq_rsp->qr);
+			  &rq_rsp->qr, "OSF");
 }

static void ufshcd_add_tm_upiu_trace(struct ufs_hba *hba, unsigned
int tag,

drivers/scsi/ufs/ufshcd.c +0 -0
@@ -340,10 +341,10 @@ static void ufshcd_add_tm_upiu_trace(struct
 ufs_hba *hba, unsigned int tag,
 

drivers/scsi/ufs/ufshcd.c static void ufshcd_add_tm_upiu_trace(struct · line 350

Quoted 10 lines
drivers/scsi/ufs/ufshcd.c +2 -2
 	if (!strcmp("tm_send", str))
 		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp-
 >req_header,
-				  &descp->input_param1);
+				  &descp->input_param1, "TM_INPUT");
 	else
 		trace_ufshcd_upiu(dev_name(hba->dev), str, &descp-
 >rsp_header,
-				  &descp->output_param1);
+				  &descp->output_param1, "TM_OUTPUT");

You could save some space on the ring buffer, if you made the above into an
enum, and then used print_symbolic().

include/trace/events/ufs.h TRACE_EVENT(ufshcd_uic_command, · line 295

Quoted 10 lines

}

static void ufshcd_add_uic_command_trace(struct ufs_hba *hba,

include/trace/events/ufs.h +0 -0
index 0bd54a184391..68e8e97a9b47 100644
@@ -295,15 +295,17 @@ TRACE_EVENT(ufshcd_uic_command,
 );

You could make this:

#define TRACE_TSF_TYPES		\
	EM(CDB)			\
	EM(OSF)			\
	EM(TM_INPUT)		\
	EMe(TM_OUTPUT)

#ifndef TRACE_TSF_TYPES_ENUMS
#define TRACE_TSF_TYPES_ENUMS
#undef EM
#undef EMe

#define EM(x) TRACE_TSF_##x,
#define EMe(x) TRACE_TSF_##x

enum {
TRACE_TSF_TYPES
}
#endif /* TRACE_TSF_TYPES_ENUMS */

#undef EM
#undef EMe

/* These export the enum names to user space */
#define EM(x) TRACE_DEFINE_ENUM(TRACE_TSF_##x)
#define EMe(x) TRACE_DEFINE_ENUM(TRACE_TSF_##x)

TRACE_TSF_TYPES

#undef EM
#undef EMe

/* These are used in the print_symbolic */
#define EM(x) { TRACE_TSF_##x, #x },
#define EMe(x) { TRACE_TSF_##x, #x }

include/trace/events/ufs.h TRACE_EVENT(ufshcd_uic_command, · line 300

Quoted 7 lines
include/trace/events/ufs.h +2 -1
 TRACE_EVENT(ufshcd_upiu,
-	TP_PROTO(const char *dev_name, const char *str, void *hdr, void
 *tsf),
+	TP_PROTO(const char *dev_name, const char *str, void *hdr, void
 *tsf,
+		 const char *tsf_type),
		int tsf_type;

include/trace/events/ufs.h TRACE_EVENT(ufshcd_uic_command, · line 308

Quoted 10 lines
include/trace/events/ufs.h +2 -1
-	TP_ARGS(dev_name, str, hdr, tsf),
+	TP_ARGS(dev_name, str, hdr, tsf, tsf_type),
 
 	TP_STRUCT__entry(
 		__string(dev_name, dev_name)
 		__string(str, str)
 		__array(unsigned char, hdr, 12)
 		__array(unsigned char, tsf, 16)
+		__string(tsf_type, tsf_type)
		__field(int, tsf_type)

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 316

Quoted 8 lines
include/trace/events/ufs.h +1 -0
 	),
 
 	TP_fast_assign(
@@ -311,12 +313,14 @@ TRACE_EVENT(ufshcd_upiu,
 		__assign_str(str, str);
 		memcpy(__entry->hdr, hdr, sizeof(__entry->hdr));
 		memcpy(__entry->tsf, tsf, sizeof(__entry->tsf));
+		__assign_str(tsf_type, tsf_type);
		__entry->tsf_type = tsf_type;

include/trace/events/ufs.h TRACE_EVENT(ufshcd_upiu, · line 323

Quoted 8 lines
include/trace/events/ufs.h +2 -1
 	),
 
 	TP_printk(
-		"%s: %s: HDR:%s, CDB:%s",
+		"%s: %s: HDR:%s, %s:%s",
 		__get_str(str), __get_str(dev_name),
 		__print_hex(__entry->hdr, sizeof(__entry->hdr)),
+		__get_str(tsf_type),

print_symbolic(tsf_type, TRACE_TSF_TYPES),

Quoted 8 lines

-- Steve

Quoted 3 lines
include/trace/events/ufs.h +0 -0
 		__print_hex(__entry->tsf, sizeof(__entry->tsf))
 	)
 );

Thanks Bart and Steve,
That's nice, I will change them in the next version.

Bean