Messages in this thread | | | Date | Mon, 2 Mar 2009 10:41:52 -0500 (EST) | From | Steven Rostedt <> | Subject | Re: [ANNOUNCE][PATCH 0 of 4] zedtrace, a general-purpose binary tracer |
| |
On Sat, 28 Feb 2009, Christoph Hellwig wrote:
> I played with it a bit and I really like the userspace with the > transport backend and a hakcy script language that many people know > for post-processing. > > But looking at the kernel portions just shows again that the whole > tracepoint concept is utterly wrong. We should not need the > zed_block.c/zed_sched.c/zed_workqueue.c but instead have those > tracing points directly annotated in the subsystems like using > markers, so that's it's self-contained and can be added by modules > without touching the rest of the kernel. > > I know tracepoints are not your fault, but we really need to sort > that out for a nice tracing experience.
Christoph,
Have you seen my udpates to the TRACE_EVENT_FORMAT? I based my work off of Tom's work here. That is, I liked the way he could do the C style recording of events, without the printf parsing.
The difference in my work is that I put all of it into a self contained macro called TRACE_EVENT_FORMAT:
For example:
TRACE_EVENT_FORMAT(sched_switch, TPPROTO(struct rq *rq, struct task_struct *prev, struct task_struct *next), TPARGS(rq, prev, next), TPFMT("task %s:%d ==> %s:%d", prev->comm, prev->pid, next->comm, next->pid), TRACE_STRUCT( TRACE_FIELD(pid_t, prev_pid, prev->pid) TRACE_FIELD(int, prev_prio, prev->prio) TRACE_FIELD_SPECIAL(char next_comm[TASK_COMM_LEN], next_comm, TPCMD(memcpy(TRACE_ENTRY->next_comm, next->comm, TASK_COMM_LEN))) TRACE_FIELD(pid_t, next_pid, next->pid) TRACE_FIELD(int, next_prio, next->prio) ), TPRAWFMT("prev %d:%d ==> next %s:%d:%d") );
I explained all these fields in a prior email, but I added the new TRACE_FIELD_SPECIAL here (have not posted it yet).
This one allows you to do something special if a simple assignment is not enough.
I used next_comm as an example:
TRACE_FIELD_SPECIAL(type_item, item, cmd)
Where type_item is a full line in a structure (not just the time, but the item too). This allows to allocate arrays. The item is the same as the item in the type_item, but just the variable name and not the type. The cmd is the command to perform to record it. Note, you need to use TRACE_ENTRY to access the field.
The last TRPRAWFMT is the trace point raw format that corresponds to the structure.
-- Steve
| |