Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756976AbaKUAnZ (ORCPT ); Thu, 20 Nov 2014 19:43:25 -0500 Received: from mga14.intel.com ([192.55.52.115]:26782 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755788AbaKUAnV convert rfc822-to-8bit (ORCPT ); Thu, 20 Nov 2014 19:43:21 -0500 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.07,426,1413270000"; d="scan'208";a="625911922" From: "Zheng, Lv" To: "Kirill A. Shutemov" , Peter Zijlstra , "Ingo Molnar (mingo@redhat.com)" CC: "Rafael J. Wysocki" , "Wysocki, Rafael J" , "Brown, Len" , Lv Zheng , "linux-kernel@vger.kernel.org" , "linux-acpi@vger.kernel.org" Subject: RE: [PATCH 1/6] ACPI/EC: Introduce STARTED/STOPPED flags to replace BLOCKED flag. Thread-Topic: [PATCH 1/6] ACPI/EC: Introduce STARTED/STOPPED flags to replace BLOCKED flag. Thread-Index: AQHP9yVTyINIm6IIIUqyDtdpynBt3JxRS4XwgBarhsKAAOm2sIAAvi4AgAC134A= Date: Fri, 21 Nov 2014 00:42:55 +0000 Message-ID: <1AE640813FDE7649BE1B193DEA596E88026A4128@SHSMSX101.ccr.corp.intel.com> References: <1AE640813FDE7649BE1B193DEA596E8802689778@SHSMSX101.ccr.corp.intel.com> <20141118132328.GA27428@node.dhcp.inet.fi> <4486101.LWS7CexbAj@vostro.rjw.lan> <20141119121615.GA2514@node.dhcp.inet.fi> <1AE640813FDE7649BE1B193DEA596E88026A3D69@SHSMSX101.ccr.corp.intel.com> <20141120213348.GA13221@node.dhcp.inet.fi> In-Reply-To: <20141120213348.GA13221@node.dhcp.inet.fi> Accept-Language: zh-CN, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-originating-ip: [10.239.127.40] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8BIT MIME-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, Shutemov > From: Kirill A. Shutemov [mailto:kirill@shutemov.name] > Sent: Friday, November 21, 2014 5:34 AM > > On Thu, Nov 20, 2014 at 02:20:53AM +0000, Zheng, Lv wrote: > > Since you have environment to trigger this. > > Could you also help to check if the fix can work? > > I've just sent them as RFC to this thread. > > With these two patchse on top of my -next snapshot I still see the issue: > > [ 0.324119] ====================================================== > [ 0.324125] [ INFO: possible circular locking dependency detected ] > [ 0.324132] 3.18.0-rc5-next-20141119-07477-g4c45e54745b2 #80 Not tainted > [ 0.324138] ------------------------------------------------------- > [ 0.324144] swapper/3/0 is trying to acquire lock: > [ 0.324149] (&(&ec->lock)->rlock){-.....}, at: [] acpi_ec_gpe_handler+0x21/0xfc > [ 0.324165] > but task is already holding lock: > [ 0.324171] (&(*(&acpi_gbl_gpe_lock))->rlock){-.....}, at: [] acpi_os_acquire_lock+0xe/0x10 > [ 0.324185] > which lock already depends on the new lock. > > [ 0.324193] > the existing dependency chain (in reverse order) is: > [ 0.324200] > -> #1 (&(*(&acpi_gbl_gpe_lock))->rlock){-.....}: > [ 0.324209] [] lock_acquire+0xdf/0x2d0 > [ 0.324218] [] _raw_spin_lock_irqsave+0x50/0x70 > [ 0.324228] [] acpi_os_acquire_lock+0xe/0x10 > [ 0.324235] [] acpi_enable_gpe+0x27/0x75 > [ 0.324244] [] acpi_ec_start+0x67/0x88 > [ 0.324251] [] ec_install_handlers+0x41/0xa4 > [ 0.324258] [] acpi_ec_ecdt_probe+0x1a9/0x1ea > [ 0.324267] [] acpi_init+0x8b/0x26e > [ 0.324275] [] do_one_initcall+0xd8/0x210 > [ 0.324283] [] kernel_init_freeable+0x1f5/0x282 > [ 0.324293] [] kernel_init+0xe/0xf0 > [ 0.324300] [] ret_from_fork+0x7c/0xb0 > [ 0.324307] > -> #0 (&(&ec->lock)->rlock){-.....}: > [ 0.324315] [] __lock_acquire+0x210f/0x2220 > [ 0.324323] [] lock_acquire+0xdf/0x2d0 > [ 0.324330] [] _raw_spin_lock_irqsave+0x50/0x70 > [ 0.324338] [] acpi_ec_gpe_handler+0x21/0xfc > [ 0.324346] [] acpi_ev_gpe_dispatch+0xb9/0x12e > [ 0.324353] [] acpi_ev_gpe_detect+0x105/0x227 > [ 0.324360] [] acpi_ev_sci_xrupt_handler+0x22/0x38 > [ 0.324368] [] acpi_irq+0x16/0x31 > [ 0.324375] [] handle_irq_event_percpu+0x6f/0x540 > [ 0.324384] [] handle_irq_event+0x41/0x70 > [ 0.324392] [] handle_fasteoi_irq+0x86/0x140 > [ 0.324399] [] handle_irq+0x22/0x40 > [ 0.324408] [] do_IRQ+0x4f/0xf0 > [ 0.324416] [] ret_from_intr+0x0/0x1a > [ 0.324423] [] default_idle+0x23/0x260 > [ 0.324430] [] arch_cpu_idle+0xf/0x20 > [ 0.324438] [] cpu_startup_entry+0x36b/0x5b0 > [ 0.324445] [] start_secondary+0x1a4/0x1d0 > [ 0.324454] > other info that might help us debug this: > > [ 0.324462] Possible unsafe locking scenario: > > [ 0.324468] CPU0 CPU1 > [ 0.324473] ---- ---- > [ 0.324477] lock(&(*(&acpi_gbl_gpe_lock))->rlock); > [ 0.324483] lock(&(&ec->lock)->rlock); > [ 0.324490] lock(&(*(&acpi_gbl_gpe_lock))->rlock); > [ 0.324498] lock(&(&ec->lock)->rlock); > [ 0.324503] Let me convert this into call stack: CPU0 CPU1 acpi_irq +GPE acpi_ev_gpe_dispatch acpi_bus_init acpi_ec_ecdt_probe acpi_install_gpe_handler() +EC acpi_ec_start +GPE acpi_enable_gpe -GPE -EC +EC acpi_ec_gpe_handler -EC -GPE I used + to indicate spin_lock() and - to indicate spin_unlock(). GPE to indicate acpi_gbl_gpe_lock, EC to indicate ec->lock. Are you sure you still can see this? Please help to check the [RFC PATCH 2] to see if the following code is exactly applied: + /* + * There is no protection around the namespace node + * and the GPE handler to ensure a safe destruction + * because: + * 1. The namespace node is expected to always + * exist after loading a table. + * 2. The GPE handler is expected to be flushed by + * acpi_os_wait_events_complete() before the + * destruction. + */ + acpi_os_release_lock + (acpi_gbl_gpe_lock, flags); + int_status |= + gpe_handler_info-> + address(gpe_device, + gpe_number, + gpe_handler_info-> + context); This is where acpi_ec_gpe_handler() will be invoked. + flags = + acpi_os_acquire_lock + (acpi_gbl_gpe_lock); So when acpi_ec_gpe_handler() is invoked, GPE lock is release. There should be no reason you can see this warning, because the call stack will be: CPU0 CPU1 CPU0 CPU1 acpi_irq +GPE acpi_ev_gpe_dispatch acpi_bus_init acpi_ec_ecdt_probe acpi_install_gpe_handler() +EC acpi_ec_start +GPE acpi_enable_gpe -GPE -EC -GPE +EC acpi_ec_gpe_handler -EC +GPE -GPE When acpi_ec_gpe_handler() is invoked, there is no acpi_gbl_gpe_lock locked. So I really cannot understand your test result. Could you confirm this again? Maybe I just don't understand how this warning is generated, and this is just a kind of warning that we can ignore. Let me ask Peter and Ingo to check if this is just a limitation of lockdep. Thanks and best regards -Lv > *** DEADLOCK *** > > [ 0.324510] 1 lock held by swapper/3/0: > [ 0.324514] #0: (&(*(&acpi_gbl_gpe_lock))->rlock){-.....}, at: [] acpi_os_acquire_lock+0xe/0x10 > [ 0.324528] > stack backtrace: > [ 0.324535] CPU: 3 PID: 0 Comm: swapper/3 Not tainted 3.18.0-rc5-next-20141119-07477-g4c45e54745b2 #80 > [ 0.324543] Hardware name: LENOVO 3460CC6/3460CC6, BIOS G6ET93WW (2.53 ) 02/04/2013 > [ 0.324550] ffffffff82cae120 ffff88011e2c3ba8 ffffffff81af484e 0000000000000011 > [ 0.324560] ffffffff82cae120 ffff88011e2c3bf8 ffffffff81af3361 0000000000000001 > [ 0.324569] ffff88011e2c3c58 ffff88011e2c3bf8 ffff8801193f92b0 ffff8801193f9b00 > [ 0.324579] Call Trace: > [ 0.324582] [] dump_stack+0x4c/0x6e > [ 0.324593] [] print_circular_bug+0x2b2/0x2c3 > [ 0.324601] [] __lock_acquire+0x210f/0x2220 > [ 0.324609] [] lock_acquire+0xdf/0x2d0 > [ 0.324616] [] ? acpi_ec_gpe_handler+0x21/0xfc > [ 0.324624] [] _raw_spin_lock_irqsave+0x50/0x70 > [ 0.324631] [] ? acpi_ec_gpe_handler+0x21/0xfc > [ 0.324640] [] ? acpi_hw_write+0x4b/0x52 > [ 0.324646] [] acpi_ec_gpe_handler+0x21/0xfc > [ 0.324653] [] acpi_ev_gpe_dispatch+0xb9/0x12e > [ 0.324660] [] acpi_ev_gpe_detect+0x105/0x227 > [ 0.324668] [] acpi_ev_sci_xrupt_handler+0x22/0x38 > [ 0.324675] [] acpi_irq+0x16/0x31 > [ 0.324683] [] handle_irq_event_percpu+0x6f/0x540 > [ 0.324691] [] handle_irq_event+0x41/0x70 > [ 0.324698] [] ? handle_fasteoi_irq+0x28/0x140 > [ 0.324705] [] handle_fasteoi_irq+0x86/0x140 > [ 0.324712] [] handle_irq+0x22/0x40 > [ 0.324719] [] do_IRQ+0x4f/0xf0 > [ 0.324725] [] common_interrupt+0x72/0x72 > [ 0.324731] [] ? native_safe_halt+0x6/0x10 > [ 0.324743] [] ? trace_hardirqs_on+0xd/0x10 > [ 0.324750] [] default_idle+0x23/0x260 > [ 0.324757] [] arch_cpu_idle+0xf/0x20 > [ 0.324763] [] cpu_startup_entry+0x36b/0x5b0 > [ 0.324771] [] start_secondary+0x1a4/0x1d0 > > > > > > Thanks and best regards > > -Lv > > > > > From: Kirill A. Shutemov [mailto:kirill@shutemov.name] > > > Sent: Wednesday, November 19, 2014 8:16 PM > > > To: Rafael J. Wysocki > > > Cc: Zheng, Lv; Wysocki, Rafael J; Brown, Len; Lv Zheng; linux-kernel@vger.kernel.org; linux-acpi@vger.kernel.org > > > Subject: Re: [PATCH 1/6] ACPI/EC: Introduce STARTED/STOPPED flags to replace BLOCKED flag. > > > > > > On Tue, Nov 18, 2014 at 10:20:11PM +0100, Rafael J. Wysocki wrote: > > > > On Tuesday, November 18, 2014 03:23:28 PM Kirill A. Shutemov wrote: > > > > > On Wed, Nov 05, 2014 at 02:52:36AM +0000, Zheng, Lv wrote: > > > > > > > > [cut] > > > > > > > > > > > > > > Here's lockdep warning I see on -next: > > > > > > > > Is patch [1/6] sufficient to trigger this or do you need all [1-4/6]? > > > > > > I only saw it on -next. I've tried to apply patches directly on -rc5 and > > > don't see the warning. I don't have time for proper bisecting, sorry. > > > > > > -- > > > Kirill A. Shutemov > > -- > Kirill A. Shutemov -- 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/