[DTrace-devel] [PATCH v2] test: clean up orphaned tracing events between tests

[Kris Van Hees]

On Wed, Jun 07, 2023 at 01:22:11PM -0400, Eugene Loh via DTrace-devel wrote:
> Reviewed-by: Eugene Loh <eugene.loh at oracle.com>
> 
> But I guess I'm curious why EVENTS is used inside the awk script while
> KPROBES and UPROBES are not;  they are "hardwired" inside the awk script to
> values that are already defined.  In case the issue is defining more "awk
> -v" variable values, one way around that (including for the existing "-v
> tracefs") is to use stuff like:
>         awk 'END { fn = '$KPROBES'; print >> fn; }'

True, I can just put in KPROBES and UPROBES.  I was using the env names mostly
to make things more readable so functionally it doesn't really matter.  But I
will make pass them in.

> With regards to argument lists for user-defined functions, the awk man page
> says "The convention is to separate local variables from real parameters by
> extra spaces in the parameter list."

I know, but I don't think that actually makes much sense (and looks kind of
odd).  It also seems to depend on which AWK implementation you prefer (some
do not really suggest any conventions).

> Should the /^===/ clause end with "next;"?

Yes, might as well (and it is cleaner).

> A few comments in the awk would be nice.  And maybe a more descriptive name
> for "re"?

re is quite common as a name for a regular expression.

> Are there dt_pid events that need to be cleaned up?

We can't because they are not identifiable by pid of the dtrace that created
them.

> On 6/7/23 12:32, Kris Van Hees via DTrace-devel wrote:
> > diff --git a/test/utils/clean_probes.sh b/test/utils/clean_probes.sh
> > +#!/usr/bin/bash
> > +
> > +TRACEFS=/sys/kernel/debug/tracing
> > +EVENTS=${TRACEFS}/available_events
> > +KPROBES=${TRACEFS}/kprobe_events
> > +UPROBES=${TRACEFS}/uprobe_events
> > +
> > +# Check permissions
> > +if [[ ! -r ${EVENTS} ]]; then
> > +	echo "ERROR: Cannot read ${EVENTS}" > /dev/stderr
> > +	exit 1
> > +elif [[ ! -w ${KPROBES} ]]; then
> > +	echo "ERROR: Cannot write to ${KPROBES}" > /dev/stderr
> > +	exit 1
> > +elif [[ ! -w ${UPROBES} ]]; then
> > +	echo "ERROR: Cannot write to ${UPROBES}" > /dev/stderr
> > +	exit 1
> > +fi
> > +
> > +# Scan the list of events for any orphaned DTrace probes
> > +{
> > +	ps -C dtrace -o pid=
> > +	echo '==='
> > +	cat ${EVENTS}
> > +} | \
> > +	awk -v tracefs=${TRACEFS} \
> > +	    'function getTimestamp(dt) {
> > +		 cmd = "date +\"%s.%N\"";
> > +		 cmd | getline dt;
> > +		 close(cmd);
> > +		 return dt;
> > +	     }
> > +
> > +	     function timeDiff(t0, t1, s0, n0, s1, n1) {
> > +		 tmp = $0;
> > +
> > +		 $0 = t0;
> > +		 sub(/\./, " ");
> > +		 s0 = int($1);
> > +		 n0 = int($2);
> > +		 $0 = t1;
> > +		 sub(/\./, " ");
> > +		 s1 = int($1);
> > +		 n1 = int($2);
> > +
> > +		 if (n1 < n0) {
> > +		     s1--;
> > +		     n1 += 1000000000;
> > +		 }
> > +
> > +		 $0 = tmp;
> > +
> > +		 return sprintf("%d.%09d", s1 - s0, n1 - n0);
> > +	     }
> > +
> > +	     BEGIN { start = getTimestamp(); }
> > +
> > +	     /^===/ { re = "^dt_(" substr(pids, 2) ")_"; }
> > +
> > +	     !re {
> > +		 pids = pids "|" int($1);
> > +		 next;
> > +	     }
> > +
> > +	     $1 !~ /^dt_[0-9]/ { next; }
> > +	     $1 ~ re { next; }
> > +	     { sub(/:/, "/"); }
> > +
> > +	     {
> > +		 if (/_fbt_/)
> > +		     kpv[kpc++] = $0;
> > +		 else
> > +		     upv[upc++] = $0;
> > +	     }
> > +
> > +	     END {
> > +		 fn = tracefs"/kprobe_events";
> > +		 for (i = 0; i < kpc; i++) {
> > +		     print "Orphaned kprobe "kpv[i];
> > +		     print "-:"kpv[i] >> fn;
> > +		     if (i % 20 == 19)
> > +			 close(fn);
> > +		 }
> > +		 close(fn);
> > +
> > +		 fn = tracefs"/uprobe_events";
> > +		 for (i = 0; i < upc; i++) {
> > +		     print "Orphaned uprobe "upv[i];
> > +		     print "-:"upv[i] >> fn;
> > +		     if (i % 20 == 19)
> > +			 close(fn);
> > +		 }
> > +		 close(fn);
> > +
> > +		 diff = timeDiff(start, getTimestamp());
> > +		 txt = sprintf(" (%d kprobes, %d uprobes)", kpc, upc);
> > +		 print "Cleaning took " diff txt;
> > +	     }'
> > +
> > +exit 0
> 
> _______________________________________________
> DTrace-devel mailing list
> DTrace-devel at oss.oracle.com
> https://oss.oracle.com/mailman/listinfo/dtrace-devel