Skip to contentSkip to conversation

tracing: fix "gfp_t" format for synthetic events

PATCH Superseded Zhengjun Xing 2 participants 2 messages Oct 17, 2019 Tracing

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tracing: fix "gfp_t" format for synthetic events

In the format of synthetic events, the "gfp_t" is shown as "signed:1", but in fact the "gfp_t" is "unsigned", should be shown as "signed:0". The offset should be increased by the real size of each field, rather than by the size of "u64".

The issue can be reproduced by the following commands:

echo 'memlatency u64 lat; unsigned int order; gfp_t gfp_flags; int migratetype' > /sys/kernel/debug/tracing/synthetic_events cat /sys/kernel/debug/tracing/events/synthetic/memlatency/format

name: memlatency
ID: 2233
format:

        field:unsigned short common_type;       offset:0;       size:2; signed:0;
        field:unsigned char common_flags;       offset:2;       size:1; signed:0;
        field:unsigned char common_preempt_count;       offset:3;       size:1; signed:0;
        field:int common_pid;   offset:4;       size:4; signed:1;

        field:u64 lat;  offset:8;       size:8; signed:0;
        field:unsigned int order;       offset:16;      size:4; signed:0;
        field:gfp_t gfp_flags;  offset:24;      size:4; signed:1;
        field:int migratetype;  offset:32;      size:4; signed:1;

print fmt: "lat=%llu, order=%u, gfp_flags=%x, migratetype=%d", REC->lat, REC->order, REC->gfp_flags, REC->migratetype

---
 kernel/trace/trace_events_hist.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/trace/trace_events_hist.c +3 -1
index 57648c5aa679..7d70321d03b1 100644
@@ -665,7 +665,7 @@ static int synth_event_define_fields(struct trace_event_call *call)
 			offset += STR_VAR_LEN_MAX;
 			n_u64 += STR_VAR_LEN_MAX / sizeof(u64);
 		} else {
-			offset += sizeof(u64);
+			offset += size;
 			n_u64++;
 		}
 	}
@@ -679,6 +679,8 @@ static bool synth_field_signed(char *type)
 {
 	if (str_has_prefix(type, "u"))
 		return false;
+	if (strcmp(type, "gfp_t") == 0)
+		return false;
 
 	return true;
 }
Signature · 1 line
2.17.1
#1 lore raw

Hi Zhengjun,

Quoted 51 lines · Zhengjun Xing

In the format of synthetic events, the "gfp_t" is shown as "signed:1",
but in fact the "gfp_t" is "unsigned", should be shown as "signed:0". The offset should be increased by the real size of each field, rather than by the size of "u64".

The issue can be reproduced by the following commands:

echo 'memlatency u64 lat; unsigned int order; gfp_t gfp_flags; int migratetype' > /sys/kernel/debug/tracing/synthetic_events cat /sys/kernel/debug/tracing/events/synthetic/memlatency/format

name: memlatency
ID: 2233
format:
field:unsigned short
common_type; offset:0; size:2; signed:0;
field:unsigned char
common_flags; offset:2; size:1; signed:0;
field:unsigned char

common_preempt_count;       offset:3;       size:1; signed:0;
        field:int common_pid;   offset:4;       size:4; signed:1;

        field:u64 lat;  offset:8;       size:8; signed:0;
        field:unsigned int order;       offset:16;      size:4;
signed:0;
        field:gfp_t gfp_flags;  offset:24;      size:4; signed:1;
        field:int migratetype;  offset:32;      size:4; signed:1;

print fmt: "lat=%llu, order=%u, gfp_flags=%x, migratetype=%d", REC-

Quoted 1 line

lat, REC->order, REC->gfp_flags, REC->migratetype

---
 kernel/trace/trace_events_hist.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/trace/trace_events_hist.c +1 -1
index 57648c5aa679..7d70321d03b1 100644
@@ -665,7 +665,7 @@ static int synth_event_define_fields(struct
 trace_event_call *call)
 			offset += STR_VAR_LEN_MAX;
 			n_u64 += STR_VAR_LEN_MAX / sizeof(u64);
 		} else {
-			offset += sizeof(u64);
+			offset += size;
 			n_u64++;
 		}
Quoted 1 line

}

This part isn't correct - currently, all the synthetic event fields are u64, so doing this alone will mess things up. The synthetic fields were defined to be u64 in order to simplify event generation - if we want to use the actual sizes, it would take a more extensive patch including the generation code as well. Not sure the added complexity is worth it to save a few bytes in the ring buffer.

kernel/trace/trace_events_hist.c static bool synth_field_signed(char *type) · line 686

Quoted 9 lines
kernel/trace/trace_events_hist.c +2 -0
@@ -679,6 +679,8 @@ static bool synth_field_signed(char *type)
 {
 	if (str_has_prefix(type, "u"))
 		return false;
+	if (strcmp(type, "gfp_t") == 0)
+		return false;
 
 	return true;
 }

This part is fine and makes sense.

Thanks,

Tom