From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 3308CC433EF for ; Sat, 9 Oct 2021 03:52:09 +0000 (UTC) Received: from bombadil.infradead.org (bombadil.infradead.org [198.137.202.133]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id DF4F960F0F for ; Sat, 9 Oct 2021 03:52:08 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.1 mail.kernel.org DF4F960F0F Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=gmail.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=lists.infradead.org DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=bombadil.20210309; h=Sender: Content-Transfer-Encoding:Content-Type:List-Subscribe:List-Help:List-Post: List-Archive:List-Unsubscribe:List-Id:In-Reply-To:MIME-Version:References: Message-ID:Subject:Cc:To:From:Date:Reply-To:Content-ID:Content-Description: Resent-Date:Resent-From:Resent-Sender:Resent-To:Resent-Cc:Resent-Message-ID: List-Owner; bh=AO+6vlC84tq6zU3W+dbfY3eqx/KwEZcl/LLmw7sYTKw=; b=aGqCSpuU1WuKyX 7kApnD5gY/PP+6uGDg0XvhgaaXqVdUqiPXjMXNez+N+3axzwu81x01O8xWM4R+SAJD0awMM2hQrhm CwrA/FqHpyOrGDfLrCwgTjEN87yc+nD3BsFibhWk7OAVF1WATi272KHR5XLD6EL+1//3mkfo778rP xDvU5hkXXaAVQfwmz3aoBEGrUobLOjKyRlF4KvRyvwcOfQPgjmeAFj4BiVO224QwRsfWwS0Z8k6nT 6P0Bz0yIBQqUkR461rzJsFWGSaeivNl95D7mu47q5NINgYGYXa9gQsGPaXwVWeZUHnkbV4OM6AoPQ vFy9JMVsCcVb6BvZ1rtw==; Received: from localhost ([::1] helo=bombadil.infradead.org) by bombadil.infradead.org with esmtp (Exim 4.94.2 #2 (Red Hat Linux)) id 1mZ3N3-0052f8-2Q; Sat, 09 Oct 2021 03:50:01 +0000 Received: from mail-pl1-x636.google.com ([2607:f8b0:4864:20::636]) by bombadil.infradead.org with esmtps (Exim 4.94.2 #2 (Red Hat Linux)) id 1mZ3Mx-0052e3-Cx for linux-arm-kernel@lists.infradead.org; Sat, 09 Oct 2021 03:49:57 +0000 Received: by mail-pl1-x636.google.com with SMTP id w14so7416887pll.2 for ; Fri, 08 Oct 2021 20:49:55 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=date:from:to:cc:subject:message-id:references:mime-version :content-disposition:in-reply-to; bh=YsOo6uJpHFuwAQVn6OjX/o2BoeDbNko2xmVe2QMnXMU=; b=jB+fqNCISS9Dpu/l3tmqlCxrHSjotS0mWBWaYn1dhjl5Pq5AJ3I0csA17iVRNhsAfH M/ucWx8hzQBGsqktvMsGjrHWwL0GPNxApjbq+Lkww2RmLsz2S4orK49qYZy52YAoiBB+ yfongPR4yguxoG76fp3l4EStXuKCuTfb5dRkpBPGphd8Jnd8CZf5aev38CbJQJW5dqmg bbSWgJ6D2pnt2KgHCdBFudgvZM3V73f0HpjNsBS9LwpYaH9S2Dqa/uO9k5V1c/ODgZ9F rh2WxFiSnohp1CxHcppL4uuBmFzQ0VWpfyuj/XStZ4v93a2dR/E7UA5q9OGORKT4NwN6 F+bQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:date:from:to:cc:subject:message-id:references :mime-version:content-disposition:in-reply-to; bh=YsOo6uJpHFuwAQVn6OjX/o2BoeDbNko2xmVe2QMnXMU=; b=QyQ1WnzxAeh1ycAQ7NUmCDJHEbgGjj8BvTKNNjZbtoBQJc+gXbH3FpXSKrwWrnHkDK Cb6EQQ3+aVm8c5VK2fVA6APH6RO9rXqUM5hZfmWUeDz0xQ0UO5SL2POuKKvJgezrPtsW LYx910mo2VgkWlC/ij8EO+jm1IVQkBNMUkc9zVPq0dCSGDmt2rxOzBT5TVyH+p1QNxQ9 pmei91B8T+KLJ7FXfjgnxsEyvOOVtKR6QrY3n0s8F/CJD2g8kGj+a9M+Q3Puv1iN6vUD q9MzfIBMh0SBC2OaAF2jBGVBX9ECavcosBcD341mWsYRZTRMYGLVoPhItOYV4Hhv/0/3 uqkw== X-Gm-Message-State: AOAM530I1O54XoUDA+r+Syx/B3spFALSzSH0eSX+csP8EwVs43n/RoLz LMeo+jPAc4ralN2ao5l3jA== X-Google-Smtp-Source: ABdhPJxQZXlwH2PohsiZXf9412JVOod24wFAdxtPoy7iglEJeotIjVMAdw2Gy4ELeqU2x1whFg44Ug== X-Received: by 2002:a17:902:7d89:b0:13c:a5e1:f0f1 with SMTP id a9-20020a1709027d8900b0013ca5e1f0f1mr12865501plm.24.1633751394201; Fri, 08 Oct 2021 20:49:54 -0700 (PDT) Received: from piliu.users.ipa.redhat.com ([209.132.188.80]) by smtp.gmail.com with ESMTPSA id l14sm12853915pjq.13.2021.10.08.20.49.49 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 08 Oct 2021 20:49:53 -0700 (PDT) Date: Sat, 9 Oct 2021 11:49:47 +0800 From: Pingfan Liu To: Mark Rutland Cc: "Paul E. McKenney" , linux-arm-kernel@lists.infradead.org, Catalin Marinas , Will Deacon , Marc Zyngier , Joey Gouly , Sami Tolvanen , Julien Thierry , Thomas Gleixner , Yuichi Ito , linux-kernel@vger.kernel.org Subject: Re: [PATCHv2 1/5] arm64/entry-common: push the judgement of nmi ahead Message-ID: References: <20210924132837.45994-1-kernelfans@gmail.com> <20210924132837.45994-2-kernelfans@gmail.com> <20210924175306.GB42068@C02TD0UTHF1T.local> <20210930133257.GB18258@lakrids.cambridge.arm.com> <20211008172513.GD976@C02TD0UTHF1T.local> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <20211008172513.GD976@C02TD0UTHF1T.local> X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20211008_204955_515560_E0C0C519 X-CRM114-Status: GOOD ( 61.02 ) X-BeenThere: linux-arm-kernel@lists.infradead.org X-Mailman-Version: 2.1.34 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Sender: "linux-arm-kernel" Errors-To: linux-arm-kernel-bounces+linux-arm-kernel=archiver.kernel.org@lists.infradead.org On Fri, Oct 08, 2021 at 06:25:13PM +0100, Mark Rutland wrote: > On Fri, Oct 08, 2021 at 10:55:04PM +0800, Pingfan Liu wrote: > > On Fri, Oct 08, 2021 at 12:01:25PM +0800, Pingfan Liu wrote: > > > Sorry that I missed this message and I am just back from a long > > > festival. > > > > > > Adding Paul for RCU guidance. > > > > > > On Thu, Sep 30, 2021 at 02:32:57PM +0100, Mark Rutland wrote: > > > > On Sat, Sep 25, 2021 at 11:39:55PM +0800, Pingfan Liu wrote: > > > > > On Fri, Sep 24, 2021 at 06:53:06PM +0100, Mark Rutland wrote: > > > > > > On Fri, Sep 24, 2021 at 09:28:33PM +0800, Pingfan Liu wrote: > > > > > > > In enter_el1_irq_or_nmi(), it can be the case which NMI interrupts an > > > > > > > irq, which makes the condition !interrupts_enabled(regs) fail to detect > > > > > > > the NMI. This will cause a mistaken account for irq. > > > > > > > > > > > Sorry about the confusing word "account", it should be "lockdep/rcu/.." > > > > > > > > > > > Can you please explain this in more detail? It's not clear which > > > > > > specific case you mean when you say "NMI interrupts an irq", as that > > > > > > could mean a number of distinct scenarios. > > > > > > > > > > > > AFAICT, if we're in an IRQ handler (with NMIs unmasked), and an NMI > > > > > > causes a new exception we'll do the right thing. So either I'm missing a > > > > > > subtlety or you're describing a different scenario.. > > > > > > > > > > > > Note that the entry code is only trying to distinguish between: > > > > > > > > > > > > a) This exception is *definitely* an NMI (because regular interrupts > > > > > > were masked). > > > > > > > > > > > > b) This exception is *either* and IRQ or an NMI (and this *cannot* be > > > > > > distinguished until we acknowledge the interrupt), so we treat it as > > > > > > an IRQ for now. > > > > > > > > > > > b) is the aim. > > > > > > > > > > At the entry, enter_el1_irq_or_nmi() -> enter_from_kernel_mode()->rcu_irq_enter()/rcu_irq_enter_check_tick() etc. > > > > > While at irqchip level, gic_handle_irq()->gic_handle_nmi()->nmi_enter(), > > > > > which does not call rcu_irq_enter_check_tick(). So it is not proper to > > > > > "treat it as an IRQ for now" > > > > > > > > I'm struggling to understand the problem here. What is "not proper", and > > > > why? > > > > > > > > Do you think there's a correctness problem, or that we're doing more > > > > work than necessary? > > > > > > > I had thought it just did redundant accounting. But after revisiting RCU > > > code, I think it confronts a real bug. > > > > > > > If you could give a specific example of a problem, it would really help. > > > > > > > Refer to rcu_nmi_enter(), which can be called by > > > enter_from_kernel_mode(): > > > > > > ||noinstr void rcu_nmi_enter(void) > > > ||{ > > > || ... > > > || if (rcu_dynticks_curr_cpu_in_eqs()) { > > > || > > > || if (!in_nmi()) > > > || rcu_dynticks_task_exit(); > > > || > > > || // RCU is not watching here ... > > > || rcu_dynticks_eqs_exit(); > > > || // ... but is watching here. > > > || > > > || if (!in_nmi()) { > > > || instrumentation_begin(); > > > || rcu_cleanup_after_idle(); > > > || instrumentation_end(); > > > || } > > > || > > > || instrumentation_begin(); > > > || // instrumentation for the noinstr rcu_dynticks_curr_cpu_in_eqs() > > > || instrument_atomic_read(&rdp->dynticks, sizeof(rdp->dynticks)); > > > || // instrumentation for the noinstr rcu_dynticks_eqs_exit() > > > || instrument_atomic_write(&rdp->dynticks, sizeof(rdp->dynticks)); > > > || > > > || incby = 1; > > > || } else if (!in_nmi()) { > > > || instrumentation_begin(); > > > || rcu_irq_enter_check_tick(); > > > || } else { > > > || instrumentation_begin(); > > > || } > > > || ... > > > ||} > > > > > > > Forget to supplement the context for understanding the case: > > On arm64, at present, a pNMI (akin to NMI) may call rcu_nmi_enter() > > without calling "__preempt_count_add(NMI_OFFSET + HARDIRQ_OFFSET);". > > As a result it can be mistaken as an normal interrupt in > > rcu_nmi_enter(). > > I appreciate that there's a window where we treat the pNMI like an IRQ, > but that's by design, and we account for this in gic_handle_irq() and > gic_handle_nmi() where we "upgrade" to NMI context with > nmi_enter()..nmi_exit(). > > The idea is that we have two cases: > > 1) If we take a pNMI from a context where IRQs were masked, we know it > must be a pNMI, and perform the NMI entry immediately to avoid > reentrancy problems. > > I think we're all happy with this case. > Right. > 2) If we take a pNMI from a context where IRQs were unmasked, we don't know > whether the trigger was a pNMI/IRQ until we read from the GIC, and > since we *could* have taken an IRQ, this is equivalent to taking a > spurious IRQ, and while handling that, taking the NMI, e.g. > > < run with IRQs unmasked > > ~~~ take IRQ ~~~ > < enter IRQ > > ~~~ take NMI exception ~~~ > < enter NMI > > < handle NMI > > < exit NMI > > ~~~ return from NMI exception ~~~ > < handle IRQ / spurious / do-nothing > > < exit IRQ > > ~~~ return from IRQ exception ~~~ > < continue running with IRQs unmasked > > Yes, here I am on the same page. (I think I used a wrong example in previous email, which caused the confusion) > ... except that we don't do the HW NMI exception entry/exit, just all > the necessary SW accounting. > A little but important thing: local_irq_save() etc can not disable pNMI. > > Note that case (2) can *never* nest within itself or within case (1). > > Do you have a specific example of something that goes wrong with the > above? e.g. something that's inconsistent with that rationale? > Please see the following comment. > > And this may cause the following issue: > > > There is 3 pieces of code put under the > > > protection of if (!in_nmi()). At least the last one > > > "rcu_irq_enter_check_tick()" can trigger a hard lock up bug. Because it > > > is supposed to hold a spin lock with irqoff by > > > "raw_spin_lock_rcu_node(rdp->mynode)", but pNMI can breach it. The same > > > scenario in rcu_nmi_exit()->rcu_prepare_for_idle(). Sorry that this should be an wrong example, since here it takes the case (1). Concentrating on the spin lock rcu_node->lock, there are two operators: raw_spin_lock_rcu_node() raw_spin_trylock_rcu_node() Then suppose the scenario for deadlock: note_gp_changes() in non-irq-context { local_irq_save(flags); ... raw_spin_trylock_rcu_node(rnp) // hold lock needwake = __note_gp_changes(rnp, rdp); ------\ raw_spin_unlock_irqrestore_rcu_node(rnp, flags); \ } \ \---> pNMI breaks in due to local_irq_save() can not disable it. rcu_irq_enter() without __preempt_count_add(NMI_OFFSET + HARDIRQ_OFFSET) ->rcu_nmi_enter() { else if (!in_nmi()) rcu_irq_enter_check_tick() ->__rcu_irq_enter_check_tick() { ... raw_spin_lock_rcu_node(rdp->mynode); //Oops deadlock! } } > > > > > > As for the first two "if (!in_nmi())", I have no idea of why, except > > > breaching spin_lock_irq() by NMI. Hope Paul can give some guide. > > That code (in enter_from_kernel_mode()) only runs in case 2, where it > cannot be nested within a pNMI, so I struggle to see how this can > deadlock. It it can, then I would expect the general case of a pNMI > nesting within and IRQ would be broken? > Sorry again for the previous misleading wrong example. Hope my new example can help. > Can you give a concrete example of a sequence that would lockup? > Currently I can't see how that's possible. > It seems the RCU subsystem has a strict semantic on NMI and normal interrupt. Besides the deadlock example, there may be other supprise to confront with (will trace it on another mail with Paul) Thanks, Pingfan _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel