Commit 4bec284cc0b9 ("tracing/user_events: Use print_format_fields() for
trace output") use print_event_fields() as safe and gives user readable
output. However, due to the insertion of the struct ftrace_event_field
structure into the field linked list from the header, the trace output
oder of fields of user events is reversed. Fix the problem by insertint
to the tail of field linked list.
Signed-off-by: sunliming <[email protected]>
---
kernel/trace/trace_events_user.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/trace/trace_events_user.c b/kernel/trace/trace_events_user.c
index aacd22c1e9f8..e9e2ec3c7613 100644
--- a/kernel/trace/trace_events_user.c
+++ b/kernel/trace/trace_events_user.c
@@ -972,7 +972,7 @@ static int user_event_add_field(struct user_event *user, const char *type,
if (filter_type == FILTER_OTHER)
field->filter_type = filter_assign_type(type);
- list_add(&field->link, &user->fields);
+ list_add_tail(&field->link, &user->fields);
/*
* Min size from user writes that are required, this does not include
--
2.25.1
On Thu, May 25, 2023 at 01:40:32PM +0800, sunliming wrote:
> Commit 4bec284cc0b9 ("tracing/user_events: Use print_format_fields() for
> trace output") use print_event_fields() as safe and gives user readable
> output. However, due to the insertion of the struct ftrace_event_field
> structure into the field linked list from the header, the trace output
> oder of fields of user events is reversed. Fix the problem by insertint
> to the tail of field linked list.
>
> Signed-off-by: sunliming <[email protected]>
> ---
> kernel/trace/trace_events_user.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/trace/trace_events_user.c b/kernel/trace/trace_events_user.c
> index aacd22c1e9f8..e9e2ec3c7613 100644
> --- a/kernel/trace/trace_events_user.c
> +++ b/kernel/trace/trace_events_user.c
> @@ -972,7 +972,7 @@ static int user_event_add_field(struct user_event *user, const char *type,
> if (filter_type == FILTER_OTHER)
> field->filter_type = filter_assign_type(type);
>
> - list_add(&field->link, &user->fields);
> + list_add_tail(&field->link, &user->fields);
>
> /*
> * Min size from user writes that are required, this does not include
> --
> 2.25.1
Thanks for the patch, however, this breaks the tracefs format file. The
fields are required to be put in backwards since it walks them
backwards.
Example using this:
echo 'u:test u32 a; u32 b;' > dynamic_events
cat /sys/kernel/tracing/events/user_events/test/format
Before this change:
name: test
...
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:u32 a; offset:8; size:4; signed:0;
field:u32 b; offset:12; size:4; signed:0;
print fmt: "a=%u b=%u", REC->a, REC->b
After this change:
name: test
...
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:u32 b; offset:12; size:4; signed:0;
field:u32 a; offset:8; size:4; signed:0;
print fmt: "b=%u a=%u", REC->b, REC->a
I do agree though, that print_fields() is doing it backwards. Can you
please fix the print_fields() function instead? (It should walk the list
of fields backwards like tracefs format file does).
Steven can then Ack that work, since it's isolated there.
Thanks,
-Beau
OK，I got it.
Beau Belgrave <[email protected]> 于2023年5月26日周五 00:53写道：
>
> On Thu, May 25, 2023 at 01:40:32PM +0800, sunliming wrote:
> > Commit 4bec284cc0b9 ("tracing/user_events: Use print_format_fields() for
> > trace output") use print_event_fields() as safe and gives user readable
> > output. However, due to the insertion of the struct ftrace_event_field
> > structure into the field linked list from the header, the trace output
> > oder of fields of user events is reversed. Fix the problem by insertint
> > to the tail of field linked list.
> >
> > Signed-off-by: sunliming <[email protected]>
> > ---
> > kernel/trace/trace_events_user.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/kernel/trace/trace_events_user.c b/kernel/trace/trace_events_user.c
> > index aacd22c1e9f8..e9e2ec3c7613 100644
> > --- a/kernel/trace/trace_events_user.c
> > +++ b/kernel/trace/trace_events_user.c
> > @@ -972,7 +972,7 @@ static int user_event_add_field(struct user_event *user, const char *type,
> > if (filter_type == FILTER_OTHER)
> > field->filter_type = filter_assign_type(type);
> >
> > - list_add(&field->link, &user->fields);
> > + list_add_tail(&field->link, &user->fields);
> >
> > /*
> > * Min size from user writes that are required, this does not include
> > --
> > 2.25.1
>
> Thanks for the patch, however, this breaks the tracefs format file. The
> fields are required to be put in backwards since it walks them
> backwards.
>
> Example using this:
> echo 'u:test u32 a; u32 b;' > dynamic_events
> cat /sys/kernel/tracing/events/user_events/test/format
>
> Before this change:
> name: test
> ...
> 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:u32 a; offset:8; size:4; signed:0;
> field:u32 b; offset:12; size:4; signed:0;
>
> print fmt: "a=%u b=%u", REC->a, REC->b
>
> After this change:
> name: test
> ...
> 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:u32 b; offset:12; size:4; signed:0;
> field:u32 a; offset:8; size:4; signed:0;
>
> print fmt: "b=%u a=%u", REC->b, REC->a
>
> I do agree though, that print_fields() is doing it backwards. Can you
> please fix the print_fields() function instead? (It should walk the list
> of fields backwards like tracefs format file does).
>
> Steven can then Ack that work, since it's isolated there.
>
> Thanks,
> -Beau