[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAEf4BzZ87DAtQSKOOLjADP3C7_4FwNw6iZr_OKYtPNO=RqFAjQ@mail.gmail.com>
Date: Wed, 3 Sep 2025 11:20:01 -0700
From: Andrii Nakryiko <andrii.nakryiko@...il.com>
To: Oleg Nesterov <oleg@...hat.com>
Cc: Jiri Olsa <jolsa@...nel.org>, Masami Hiramatsu <mhiramat@...nel.org>,
Peter Zijlstra <peterz@...radead.org>, Andrii Nakryiko <andrii@...nel.org>, bpf@...r.kernel.org,
linux-kernel@...r.kernel.org, linux-trace-kernel@...r.kernel.org,
x86@...nel.org, Song Liu <songliubraving@...com>, Yonghong Song <yhs@...com>,
John Fastabend <john.fastabend@...il.com>, Hao Luo <haoluo@...gle.com>,
Steven Rostedt <rostedt@...dmis.org>, Ingo Molnar <mingo@...nel.org>
Subject: Re: [PATCH perf/core 02/11] uprobes: Skip emulate/sstep on unique
uprobe when ip is changed
On Wed, Sep 3, 2025 at 4:28 AM Oleg Nesterov <oleg@...hat.com> wrote:
>
> On 09/02, Jiri Olsa wrote:
> >
> > If user decided to take execution elsewhere, it makes little sense
> > to execute the original instruction, so let's skip it.
>
> Exactly.
>
> So why do we need all these "is_unique" complications? Only a single
I second this. This whole is_unique flag just seems like an
unnecessary thing that spills all around (extra kernel and libbpf
flags/APIs), and it's all just not to confuse the second uprobe
attached? Let's just allow uprobes to override user registers and
handle IP change on kernel side (as unlikely() check)?
> is_unique/exclusive consumer can change regs->ip, so I guess handle_swbp()
> can just do
>
> handler_chain(uprobe, regs);
> if (instruction_pointer(regs) != bp_vaddr)
> goto out;
>
>
> > Allowing this
> > behaviour only for uprobe with unique consumer attached.
>
> But if a non-exclusive consumer changes regs->ip, we have a problem
> anyway, right?
>
> We can probably add something like
>
> rc = uc->handler(uc, regs, &cookie);
> + WARN_ON(!uc->is_unique && instruction_pointer(regs) != bp_vaddr);
>
> into handler_chain(), although I don't think this is needed.
>
> Oleg.
>
> > Signed-off-by: Jiri Olsa <jolsa@...nel.org>
> > ---
> > kernel/events/uprobes.c | 13 ++++++++++---
> > 1 file changed, 10 insertions(+), 3 deletions(-)
> >
> > diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
> > index b9b088f7333a..da8291941c6b 100644
> > --- a/kernel/events/uprobes.c
> > +++ b/kernel/events/uprobes.c
> > @@ -2568,7 +2568,7 @@ static bool ignore_ret_handler(int rc)
> > return rc == UPROBE_HANDLER_REMOVE || rc == UPROBE_HANDLER_IGNORE;
> > }
> >
> > -static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs)
> > +static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs, bool *is_unique)
> > {
> > struct uprobe_consumer *uc;
> > bool has_consumers = false, remove = true;
> > @@ -2582,6 +2582,9 @@ static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs)
> > __u64 cookie = 0;
> > int rc = 0;
> >
> > + if (is_unique)
> > + *is_unique |= uc->is_unique;
> > +
> > if (uc->handler) {
> > rc = uc->handler(uc, regs, &cookie);
> > WARN(rc < 0 || rc > 2,
> > @@ -2735,6 +2738,7 @@ static void handle_swbp(struct pt_regs *regs)
> > {
> > struct uprobe *uprobe;
> > unsigned long bp_vaddr;
> > + bool is_unique = false;
> > int is_swbp;
> >
> > bp_vaddr = uprobe_get_swbp_addr(regs);
> > @@ -2789,7 +2793,10 @@ static void handle_swbp(struct pt_regs *regs)
> > if (arch_uprobe_ignore(&uprobe->arch, regs))
> > goto out;
> >
> > - handler_chain(uprobe, regs);
> > + handler_chain(uprobe, regs, &is_unique);
> > +
> > + if (is_unique && instruction_pointer(regs) != bp_vaddr)
> > + goto out;
> >
> > /* Try to optimize after first hit. */
> > arch_uprobe_optimize(&uprobe->arch, bp_vaddr);
> > @@ -2819,7 +2826,7 @@ void handle_syscall_uprobe(struct pt_regs *regs, unsigned long bp_vaddr)
> > return;
> > if (arch_uprobe_ignore(&uprobe->arch, regs))
> > return;
> > - handler_chain(uprobe, regs);
> > + handler_chain(uprobe, regs, NULL);
> > }
> >
> > /*
> > --
> > 2.51.0
> >
>
Powered by blists - more mailing lists