Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753879AbbEDMuf (ORCPT ); Mon, 4 May 2015 08:50:35 -0400 Received: from mx1.redhat.com ([209.132.183.28]:48716 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753773AbbEDMtx (ORCPT ); Mon, 4 May 2015 08:49:53 -0400 Date: Mon, 4 May 2015 14:49:26 +0200 From: Oleg Nesterov To: Ananth N Mavinakayanahalli , Anton Arapov , David Long , Denys Vlasenko , "Frank Ch. Eigler" , Ingo Molnar , Jan Willeke , Jim Keniston , Mark Wielaard , Pratyush Anand , Srikar Dronamraju Cc: linux-kernel@vger.kernel.org Subject: [PATCH 10/10] uprobes/x86: Change arch_uretprobe_is_alive() to take !chained into account Message-ID: <20150504124926.GA22529@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20150504124835.GA22462@redhat.com> User-Agent: Mutt/1.5.18 (2008-05-17) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5063 Lines: 142 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. Add the new "bool on_call" argument to arch_uretprobe_is_alive(). prepare_uretprobe()->cleanup_return_instances() path passes "true" when we know that we can do a more strict check to detect the dead frames: chained == F. In this case "sp" points to the new ret-addr, so every frame which uses the same "sp" must be dead. 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'd like to avoid another copy_from_user() especially in a case when we can't avoid the false "alive == T" positives. Signed-off-by: Oleg Nesterov --- arch/x86/kernel/uprobes.c | 10 ++++++++-- include/linux/uprobes.h | 3 ++- kernel/events/uprobes.c | 12 +++++++----- 3 files changed, 17 insertions(+), 8 deletions(-) diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c index 868dc47..f0ace39 100644 --- a/arch/x86/kernel/uprobes.c +++ b/arch/x86/kernel/uprobes.c @@ -929,7 +929,13 @@ arch_uretprobe_hijack_return_addr(struct arch_uretprobe *auret, return -1; } -bool arch_uretprobe_is_alive(struct arch_uretprobe *auret, struct pt_regs *regs) +bool arch_uretprobe_is_alive(struct arch_uretprobe *auret, + bool on_call, struct pt_regs *regs) { - return regs->sp <= auret->sp; + unsigned long sp = regs->sp; + + if (on_call) /* ->sp was just decremented by "call" insn */ + sp += sizeof_long(); + + return sp <= auret->sp; } diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h index 1ed7502..9907be4 100644 --- a/include/linux/uprobes.h +++ b/include/linux/uprobes.h @@ -129,7 +129,8 @@ extern int arch_uprobe_exception_notify(struct notifier_block *self, unsigned l extern void arch_uprobe_abort_xol(struct arch_uprobe *aup, struct pt_regs *regs); extern unsigned long arch_uretprobe_hijack_return_addr(struct arch_uretprobe *auret, unsigned long trampoline_vaddr, struct pt_regs *regs); -extern bool arch_uretprobe_is_alive(struct arch_uretprobe *auret, struct pt_regs *regs); +extern bool arch_uretprobe_is_alive(struct arch_uretprobe *auret, + bool on_call, struct pt_regs *regs); extern bool arch_uprobe_ignore(struct arch_uprobe *aup, struct pt_regs *regs); extern void arch_uprobe_copy_ixol(struct page *page, unsigned long vaddr, void *src, unsigned long len); diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c index b6433fb..ec697da 100644 --- a/kernel/events/uprobes.c +++ b/kernel/events/uprobes.c @@ -1522,10 +1522,11 @@ 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 on_call, struct pt_regs *regs) { struct return_instance *ri = utask->return_instances; - while (ri && !arch_uretprobe_is_alive(&ri->auret, regs)) { + while (ri && !arch_uretprobe_is_alive(&ri->auret, on_call, regs)) { ri = free_ret_instance(ri); utask->depth--; } @@ -1587,7 +1588,7 @@ static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs) ri->chained = chained; if (utask->depth) /* drop the entries invalidated by longjmp() */ - cleanup_return_instances(utask, regs); + cleanup_return_instances(utask, !chained, regs); utask->depth++; ri->next = utask->return_instances; @@ -1815,7 +1816,7 @@ static void handle_trampoline(struct pt_regs *regs) * could hit this trampoline on return. TODO: sigaltstack(). */ next = find_next_ret_chain(ri); - valid = !next || arch_uretprobe_is_alive(&next->auret, regs); + valid = !next || arch_uretprobe_is_alive(&next->auret, false, regs); instruction_pointer_set(regs, ri->orig_ret_vaddr); do { @@ -1840,7 +1841,8 @@ bool __weak arch_uprobe_ignore(struct arch_uprobe *aup, struct pt_regs *regs) return false; } -bool __weak arch_uretprobe_is_alive(struct arch_uretprobe *auret, struct pt_regs *regs) +bool __weak arch_uretprobe_is_alive(struct arch_uretprobe *auret, + bool on_call, struct pt_regs *regs) { return true; } -- 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/