Re: [PATCH v2 1/6] bpf: separate bvar implementation into separate functions

[Kris Van Hees <kris.van.hees@oracle.com>]

On Tue, Jan 28, 2025 at 06:29:43PM -0500, Eugene Loh wrote:
> Reviewed-by: Eugene Loh <eugene.loh@oracle.com>
> 
> If there is some quantification or other characterization of the benefit
> that could be added to the commit message, that would be nice.

Sure.

> And...
> 
> On 1/28/25 01:31, Kris Van Hees wrote:
> > The handling of builtin variables was done with a single big function
> > that therefore got linked into every single BPF program (twice if the
> > builtin variable might raise a fault).  Providing separate functions
> > for most builtin variables reduces BPF program size significantly.
> > 
> > This also reduces pressure on the BPF verifies.
> > 
> > Signed-off-by: Kris Van Hees <kris.van.hees@oracle.com>
> > 
> > diff --git a/bpf/get_bvar.c b/bpf/get_bvar.c
> > @@ -19,18 +19,21 @@
> >   # define noinline	__attribute__((noinline))
> >   #endif
> > -extern struct bpf_map_def cpuinfo;
> > -extern struct bpf_map_def probes;
> > -extern struct bpf_map_def state;
> > -extern struct bpf_map_def usdt_names;
> > -
> > -extern uint64_t PC;
> > -extern uint64_t STBSZ;
> > -extern uint64_t STKSIZ;
> > -extern uint64_t BOOTTM;
> > +extern struct bpf_map_def	cpuinfo;
> > +extern struct bpf_map_def	probes;
> > +extern struct bpf_map_def	state;
> > +extern struct bpf_map_def	usdt_names;
> > +
> > +extern uint64_t	BOOTTM;
> > +extern uint64_t	NPROBES;
> > +extern uint64_t	PC;
> > +extern uint64_t	STBSZ;
> > +extern uint64_t	STKSIZ;
> >   extern uint64_t STACK_OFF;
> > -extern uint64_t STACK_SKIP;
> > -extern uint64_t NPROBES;
> > +extern uint64_t	STACK_SKIP;
> > +extern uint64_t	TASK_COMM;
> > +extern uint64_t	TASK_REAL_PARENT;
> > +extern uint64_t	TASK_TGID;
> 
> Did you intend to change the indentation on STACK_OFF as well?

Woops, will fix.