From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756328Ab2IXPVt (ORCPT ); Mon, 24 Sep 2012 11:21:49 -0400 Received: from mx1.redhat.com ([209.132.183.28]:63266 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756271Ab2IXPVs (ORCPT ); Mon, 24 Sep 2012 11:21:48 -0400 Date: Mon, 24 Sep 2012 17:23:02 +0200 From: Oleg Nesterov To: Peter Zijlstra Cc: Ingo Molnar , Srikar Dronamraju , Ananth N Mavinakayanahalli , Anton Arapov , Sebastian Andrzej Siewior , linux-kernel@vger.kernel.org Subject: Re: [PATCH 1/4] uprobes: Kill set_swbp()->is_swbp_at_addr() Message-ID: <20120924152302.GA13098@redhat.com> References: <20120923201945.GA27446@redhat.com> <1348477426.11847.6.camel@twins> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1348477426.11847.6.camel@twins> 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 On 09/24, Peter Zijlstra wrote: > > On Sun, 2012-09-23 at 22:19 +0200, Oleg Nesterov wrote: > > A separate patch for better documentation. > > > > set_swbp()->is_swbp_at_addr() is not needed for correctness, it is > > harmless to do the unnecessary __replace_page(old_page, new_page) > > when these 2 pages are identical. > > > > And it can not be counted as optimization. mmap/register races are > > very unlikely, while in the likely case is_swbp_at_addr() adds the > > extra get_user_pages() even if the caller is uprobe_mmap(current->mm) > > and returns false. > > It does save a page of memory though... No, it doesn't, I think. If this page has int3 it is not file-backed, it was already COW'ed by uprobes or gdb. Note the !UPROBE_COPY_INSN code in install breakpoint which has another is_swbp_insn(). Yes, this logic is not 100% correct and needs more fixes. So it can only prevent the unnecessary alloc_page() + replace_page() + free_page(old_page), but only in unlikely case. And please note that 3/4 restores this optimization, but avoids the extra get_user_pages(). This will be more important when we add the filtering, uprobe_register() will need to call register_for_each_vma() every time when the new consumer comes. > > Note also that the semantics/usage of is_swbp_at_addr() in uprobe.c > > is confusing. set_swbp() uses it to detect the case when this insn > > was already modified by uprobes, that is why it should always compare > > the opcode with UPROBE_SWBP_INSN even if the hardware (like powerpc) > > has other trap insns. It doesn't matter if this "int3" was in fact > > installed by gdb or application itself, we are going to "steal" this > > breakpoint anyway and execute the original insn from vm_file even if > > it no longer matches the memory. > > > > OTOH, handle_swbp()->find_active_uprobe() uses is_swbp_at_addr() to > > figure out whether we need to send SIGTRAP or not if we can not find > > uprobe, so in this case it should return true for all trap variants, > > not only for UPROBE_SWBP_INSN. > > > > This patch removes set_swbp()->is_swbp_at_addr(), the next patches > > will remove it from set_orig_insn() which is similar to set_swbp() > > in this respect. So the only caller will be handle_swbp() and we > > can make its semantics clear. > > This does leave me with the question of _why_ you're removing it.. Again, 3/4 restores this optimization, and imho this series can be counted as a cleanup/simplification and makes sense anyway. But the main reason is dufferent. Once again. Lets ignore the problems with gdb which can install breakpoints too. set_swbp() and set_orig_insn() use is_swbp_at_addr() to figure out whether this opcode was modified by uprobes or not. So in this case is_swbp_insn() has to compare the opcode with UPROBE_SWBP_INSN used by set_swbp(). But handle_swbp()->find_active_uprobe()->is_swbp_at_addr() is different, it needs to decide should we send SIGTRAP or not if uprobe was not found. On x86 this is the same, but powerpc has other insns which can trigger powerpc's do_int3() counterpart, so in this case is_swbp_insn() should return true for all trap variants. Oleg.