On Tue, Apr 25, 2023 at 2:55 AM Jiri Olsa <olsajiri@xxxxxxxxx> wrote:
>
> On Mon, Apr 24, 2023 at 03:11:20PM -0700, Alexei Starovoitov wrote:
> > On Mon, Apr 24, 2023 at 06:04:28PM +0200, Jiri Olsa wrote:
> > > +
> > > +static int uprobe_prog_run(struct bpf_uprobe *uprobe,
> > > + unsigned long entry_ip,
> > > + struct pt_regs *regs)
> > > +{
> > > + struct bpf_uprobe_multi_link *link = uprobe->link;
> > > + struct bpf_uprobe_multi_run_ctx run_ctx = {
> > > + .entry_ip = entry_ip,
> > > + };
> > > + struct bpf_run_ctx *old_run_ctx;
> > > + int err;
> > > +
> > > + preempt_disable();
> >
> > preempt_disable? Which year is this? :)
> > Let's allow sleepable from the start.
> > See bpf_prog_run_array_sleepable.
>
> ok, we should probably add also 'multi.uprobe.s' section so the program
> gets loaded with the BPF_F_SLEEPABLE flag.. or maybe we can enable that
> by default for 'multi.uprobe' section
"uprobe.multi.s" rather, to follow "kprobe.multi.s". But we can't make
it sleepable always/by default. Sleepable BPF programs are not just
better types of programs, they have their own limitations, so it has
to be the user's choice to go with sleepable or non-sleepable.
>
> >
> > Other than this the set looks great.
> >
> > > +
> > > + if (unlikely(__this_cpu_inc_return(bpf_prog_active) != 1)) {
> > > + err = 0;
> > > + goto out;
> > > + }
> > > +
> > > + rcu_read_lock();
> > > + old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
> > > + err = bpf_prog_run(link->link.prog, regs);
> > > + bpf_reset_run_ctx(old_run_ctx);
> > > + rcu_read_unlock();
> > > +
[...]
