Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757266AbbGGNMO (ORCPT ); Tue, 7 Jul 2015 09:12:14 -0400 Received: from mail-wi0-f175.google.com ([209.85.212.175]:37206 "EHLO mail-wi0-f175.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756900AbbGGNME (ORCPT ); Tue, 7 Jul 2015 09:12:04 -0400 Date: Tue, 7 Jul 2015 15:11:58 +0200 From: Anton Arapov To: Oleg Nesterov Cc: Ananth Mavinakayanahalli , David Long , Denys Vlasenko , "Frank Ch. Eigler" , Ingo Molnar , Jan Willeke , Jim Keniston , Mark Wielaard , Pratyush Anand , Srikar Dronamraju , linux-kernel@vger.kernel.org Subject: Re: [PATCH v2 11/11] uprobes/x86: Make arch_uretprobe_is_alive(RP_CHECK_CALL) more clever Message-ID: <20150707131158.GK5946@troglodyte.intinfra.com> References: <20150707012210.GA7466@redhat.com> <20150707012313.GA7557@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20150707012313.GA7557@redhat.com> User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4997 Lines: 147 On Tue, Jul 07, 2015 at 03:23:13AM +0200, Oleg Nesterov wrote: > The previous change documents that cleanup_return_instances() can't > always detect the dead frames, the stack can grow. But there is one > special case which imho worth fixing: arch_uretprobe_is_alive() can > return true when the stack didn't actually grow, but the next "call" > insn uses the already invalidated frame. > > Test-case: > > #include > #include > > jmp_buf jmp; > int nr = 1024; > > void func_2(void) > { > if (--nr == 0) > return; > longjmp(jmp, 1); > } > > void func_1(void) > { > setjmp(jmp); > func_2(); > } > > int main(void) > { > func_1(); > return 0; > } > > If you ret-probe func_1() and func_2() prepare_uretprobe() hits the > MAX_URETPROBE_DEPTH limit and "return" from func_2() is not reported. > > When we know that the new call is not chained, we can do the more > strict check. In this case "sp" points to the new ret-addr, so every > frame which uses the same "sp" must be dead. The only complication is > that arch_uretprobe_is_alive() needs to know was it chained or not, so > we add the new RP_CHECK_CHAIN_CALL enum and change prepare_uretprobe() > to pass RP_CHECK_CALL only if !chained. > > Note: arch_uretprobe_is_alive() could also re-read *sp and check if > this word is still trampoline_vaddr. This could obviously improve the > logic, but I would like to avoid another copy_from_user() especially > in the case when we can't avoid the false "alive == T" positives. > > Signed-off-by: Oleg Nesterov Acked-by: Anton Arapov > --- > arch/x86/kernel/uprobes.c | 5 ++++- > include/linux/uprobes.h | 1 + > kernel/events/uprobes.c | 14 +++++++------- > 3 files changed, 12 insertions(+), 8 deletions(-) > > diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c > index 67eb168..a5c59f2 100644 > --- a/arch/x86/kernel/uprobes.c > +++ b/arch/x86/kernel/uprobes.c > @@ -997,5 +997,8 @@ arch_uretprobe_hijack_return_addr(unsigned long trampoline_vaddr, struct pt_regs > bool arch_uretprobe_is_alive(struct return_instance *ret, enum rp_check ctx, > struct pt_regs *regs) > { > - return regs->sp <= ret->stack; > + if (ctx == RP_CHECK_CALL) /* sp was just decremented by "call" insn */ > + return regs->sp < ret->stack; > + else > + return regs->sp <= ret->stack; > } > diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h > index c0a5402..0bdc72f 100644 > --- a/include/linux/uprobes.h > +++ b/include/linux/uprobes.h > @@ -104,6 +104,7 @@ struct return_instance { > > enum rp_check { > RP_CHECK_CALL, > + RP_CHECK_CHAIN_CALL, > RP_CHECK_RET, > }; > > diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c > index df5661a..0f370ef 100644 > --- a/kernel/events/uprobes.c > +++ b/kernel/events/uprobes.c > @@ -1511,10 +1511,11 @@ static unsigned long get_trampoline_vaddr(void) > return trampoline_vaddr; > } > > -static void cleanup_return_instances(struct uprobe_task *utask, struct pt_regs *regs) > +static void cleanup_return_instances(struct uprobe_task *utask, bool chained, > + struct pt_regs *regs) > { > struct return_instance *ri = utask->return_instances; > - enum rp_check ctx = RP_CHECK_CALL; > + enum rp_check ctx = chained ? RP_CHECK_CHAIN_CALL : RP_CHECK_CALL; > > while (ri && !arch_uretprobe_is_alive(ri, ctx, regs)) { > ri = free_ret_instance(ri); > @@ -1528,7 +1529,7 @@ static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs) > struct return_instance *ri; > struct uprobe_task *utask; > unsigned long orig_ret_vaddr, trampoline_vaddr; > - bool chained = false; > + bool chained; > > if (!get_xol_area()) > return; > @@ -1554,14 +1555,15 @@ static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs) > goto fail; > > /* drop the entries invalidated by longjmp() */ > - cleanup_return_instances(utask, regs); > + chained = (orig_ret_vaddr == trampoline_vaddr); > + cleanup_return_instances(utask, chained, regs); > > /* > * We don't want to keep trampoline address in stack, rather keep the > * original return address of first caller thru all the consequent > * instances. This also makes breakpoint unwrapping easier. > */ > - if (orig_ret_vaddr == trampoline_vaddr) { > + if (chained) { > if (!utask->return_instances) { > /* > * This situation is not possible. Likely we have an > @@ -1570,8 +1572,6 @@ static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs) > uprobe_warn(current, "handle tail call"); > goto fail; > } > - > - chained = true; > orig_ret_vaddr = utask->return_instances->orig_ret_vaddr; > } > > -- > 1.5.5.1 > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/