[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210324104058.7c06aaeb0408e24db6ba46f8@kernel.org>
Date: Wed, 24 Mar 2021 10:40:58 +0900
From: Masami Hiramatsu <mhiramat@...nel.org>
To: Peter Zijlstra <peterz@...radead.org>
Cc: Steven Rostedt <rostedt@...dmis.org>,
Ingo Molnar <mingo@...nel.org>, X86 ML <x86@...nel.org>,
Daniel Xu <dxu@...uu.xyz>, linux-kernel@...r.kernel.org,
bpf@...r.kernel.org, kuba@...nel.org, mingo@...hat.com,
ast@...nel.org, tglx@...utronix.de, kernel-team@...com, yhs@...com,
Josh Poimboeuf <jpoimboe@...hat.com>,
linux-ia64@...r.kernel.org,
Abhishek Sagar <sagar.abhishek@...il.com>
Subject: Re: [PATCH -tip v4 10/12] x86/kprobes: Push a fake return address
at kretprobe_trampoline
On Tue, 23 Mar 2021 23:30:07 +0100
Peter Zijlstra <peterz@...radead.org> wrote:
> On Mon, Mar 22, 2021 at 03:41:40PM +0900, Masami Hiramatsu wrote:
> > ".global kretprobe_trampoline\n"
> > ".type kretprobe_trampoline, @function\n"
> > "kretprobe_trampoline:\n"
> > #ifdef CONFIG_X86_64
>
> So what happens if we get an NMI here? That is, after the RET but before
> the push? Then our IP points into the trampoline but we've not done that
> push yet.
Not only NMI, but also interrupts can happen. There is no cli/sti here.
Anyway, thanks for pointing!
I think in UNWIND_HINT_TYPE_REGS and UNWIND_HINT_TYPE_REGS_PARTIAL cases
ORC unwinder also has to check the state->ip and if it is kretprobe_trampoline,
it should be recovered.
What about this?
diff --git a/arch/x86/include/asm/unwind.h b/arch/x86/include/asm/unwind.h
index 332aa6174b10..36d3971c0a2c 100644
--- a/arch/x86/include/asm/unwind.h
+++ b/arch/x86/include/asm/unwind.h
@@ -101,6 +101,15 @@ void unwind_module_init(struct module *mod, void *orc_ip, size_t orc_ip_size,
void *orc, size_t orc_size) {}
#endif
+static inline
+unsigned long unwind_recover_kretprobe(struct unwind_state *state,
+ unsigned long addr, unsigned long *addr_p)
+{
+ return is_kretprobe_trampoline(addr) ?
+ kretprobe_find_ret_addr(state->task, addr_p, &state->kr_cur) :
+ addr;
+}
+
/* Recover the return address modified by instrumentation (e.g. kretprobe) */
static inline
unsigned long unwind_recover_ret_addr(struct unwind_state *state,
@@ -110,10 +119,7 @@ unsigned long unwind_recover_ret_addr(struct unwind_state *state,
ret = ftrace_graph_ret_addr(state->task, &state->graph_idx,
addr, addr_p);
- if (is_kretprobe_trampoline(ret))
- ret = kretprobe_find_ret_addr(state->task, addr_p,
- &state->kr_cur);
- return ret;
+ return unwind_recover_kretprobe(state, ret, addr_p);
}
/*
diff --git a/arch/x86/kernel/unwind_orc.c b/arch/x86/kernel/unwind_orc.c
index 839a0698342a..cb59aeca6a4a 100644
--- a/arch/x86/kernel/unwind_orc.c
+++ b/arch/x86/kernel/unwind_orc.c
@@ -549,7 +549,15 @@ bool unwind_next_frame(struct unwind_state *state)
(void *)orig_ip);
goto err;
}
-
+ /*
+ * There is a small chance to interrupt at the entry of
+ * kretprobe_trampoline where the ORC info doesn't exist.
+ * That point is right after the RET to kretprobe_trampoline
+ * which was modified return address. So the @addr_p must
+ * be right before the regs->sp.
+ */
+ state->ip = unwind_recover_kretprobe(state, state->ip,
+ state->sp - sizeof(unsigned long));
state->regs = (struct pt_regs *)sp;
state->prev_regs = NULL;
state->full_regs = true;
@@ -562,6 +570,9 @@ bool unwind_next_frame(struct unwind_state *state)
(void *)orig_ip);
goto err;
}
+ /* See UNWIND_HINT_TYPE_REGS case comment. */
+ state->ip = unwind_recover_kretprobe(state, state->ip,
+ state->sp - sizeof(unsigned long));
if (state->full_regs)
state->prev_regs = state->regs;
--
Masami Hiramatsu <mhiramat@...nel.org>
Powered by blists - more mailing lists