Re: [PATCH v2 09/10] perf trace: Collect augmented data using BPF

[Arnaldo Carvalho de Melo <acme@kernel.org>]

On Fri, Aug 23, 2024 at 10:38:21AM -0300, Arnaldo Carvalho de Melo wrote:
> On Fri, Aug 23, 2024 at 10:24:09AM -0300, Arnaldo Carvalho de Melo wrote:
> > On Thu, Aug 15, 2024 at 09:36:25AM +0800, Howard Chu wrote:
> > > +++ b/tools/perf/util/bpf_skel/augmented_raw_syscalls.bpf.c
> > > @@ -427,7 +538,8 @@ int sys_enter(struct syscall_enter_args *args)
> > >  	 * "!raw_syscalls:unaugmented" that will just return 1 to return the
> > >  	 * unaugmented tracepoint payload.
> > >  	 */
> > > -	bpf_tail_call(args, &syscalls_sys_enter, augmented_args->args.syscall_nr);
> > > +	if (augment_sys_enter(args, &augmented_args->args))
> > > +		bpf_tail_call(args, &syscalls_sys_enter, augmented_args->args.syscall_nr);
> > 
> > We shouldn't do that, instead we keep doing
> > 
> > 	bpf_tail_call(args, &syscalls_sys_enter, augmented_args->args.syscall_nr);
> > 
> > And userspace will setup the syscalls_sys_enter map adding the generic
> > pointer collector (augment_sys_enter) for syscalls that have pointers
> > _and_ are not serviced by a pre-existing, specialized handler, this way
> > we keep the ones we have already and that already take into account
> > pretty printing network addresses based on the network family, knows how
> > to pretty print flags (the perf_event_open, etc).
> > 
> > I'll try to do this now.
> 
> So, step by step, first this, and then hook it to the syscalls that:
> 
> 1) have a pointer to collect and no handler, i.e. as a last step in
> assigning the functions to be tail called from the syscalls BPF map.
> 

Sorry, sent the wrong patch, this one is the right one:

diff --git a/tools/perf/util/bpf_skel/augmented_raw_syscalls.bpf.c b/tools/perf/util/bpf_skel/augmented_raw_syscalls.bpf.c
index f29a8dfca044649b..4c8176f9a77ca5bb 100644
--- a/tools/perf/util/bpf_skel/augmented_raw_syscalls.bpf.c
+++ b/tools/perf/util/bpf_skel/augmented_raw_syscalls.bpf.c
@@ -398,7 +398,11 @@ static bool pid_filter__has(struct pids_filtered *pids, pid_t pid)
 	return bpf_map_lookup_elem(pids, &pid) != NULL;
 }
 
-static int augment_sys_enter(void *ctx, struct syscall_enter_args *args)
+// Will be tail called for syscalls with pointers, the setup is done
+// in builtin-trace.c as the fallback for syscalls not handled by specialed code,
+// like the network ones that need to look at one field to then decide how to
+// pretty print a network specific address, etc.
+int sys_enter_augmented(struct syscall_enter_args *args)
 {
 	bool augmented, do_output = false;
 	int zero = 0, size, aug_size, index, output = 0,
@@ -480,7 +484,7 @@ static int augment_sys_enter(void *ctx, struct syscall_enter_args *args)
 	if (!do_output)
 		return 1;
 
-	return augmented__beauty_output(ctx, payload, sizeof(struct syscall_enter_args) + output);
+	return augmented__beauty_output(args, payload, sizeof(struct syscall_enter_args) + output);
 }
 
 SEC("tp/raw_syscalls/sys_enter")
@@ -511,8 +515,7 @@ int sys_enter(struct syscall_enter_args *args)
 	 * "!raw_syscalls:unaugmented" that will just return 1 to return the
 	 * unaugmented tracepoint payload.
 	 */
-	if (augment_sys_enter(args, &augmented_args->args))
-		bpf_tail_call(args, &syscalls_sys_enter, augmented_args->args.syscall_nr);
+	bpf_tail_call(args, &syscalls_sys_enter, augmented_args->args.syscall_nr);
 
 	// If not found on the PROG_ARRAY syscalls map, then we're filtering it:
 	return 0;