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 7B4A4C433F5 for ; Tue, 2 Nov 2021 15:12:40 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 5272D60F02 for ; Tue, 2 Nov 2021 15:12:40 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S234188AbhKBPPN (ORCPT ); Tue, 2 Nov 2021 11:15:13 -0400 Received: from foss.arm.com ([217.140.110.172]:37740 "EHLO foss.arm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S234652AbhKBPN4 (ORCPT ); Tue, 2 Nov 2021 11:13:56 -0400 Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.121.207.14]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id C557DD6E; Tue, 2 Nov 2021 08:11:18 -0700 (PDT) Received: from lpieralisi (e121166-lin.cambridge.arm.com [10.1.196.255]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id C913B3F7B4; Tue, 2 Nov 2021 08:11:16 -0700 (PDT) Date: Tue, 2 Nov 2021 15:11:14 +0000 From: Lorenzo Pieralisi To: "Rafael J. Wysocki" Cc: "Rafael J. Wysocki" , Mark Kettenis , Jia He , Harb Abdulhamid , Will Deacon , Len Brown , Robert Moore , Erik Kaneda , Linux ARM , Linux Kernel Mailing List , ACPI Devel Maling List , "open list:ACPI COMPONENT ARCHITECTURE (ACPICA)" , Ard Biesheuvel , Hanjun Guo , Catalin Marinas Subject: Re: [PATCH v2] Revert "ACPI: Add memory semantics to acpi_os_map_memory()" Message-ID: <20211102151114.GB15962@lpieralisi> References: <20210910122820.26886-1-justin.he@arm.com> <20210910143223.6705-1-justin.he@arm.com> <20210922163336.GA24633@lpieralisi> <56147a0b8b9fba46@bloch.sibelius.xs4all.nl> <20210923094031.GA6454@lpieralisi> <56147c6e73afe9f6@bloch.sibelius.xs4all.nl> <20210924090409.GA26089@lpieralisi> <44f20f07-257b-a1a5-23d9-ffd66bf45887@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <44f20f07-257b-a1a5-23d9-ffd66bf45887@intel.com> User-Agent: Mutt/1.9.4 (2018-02-28) Precedence: bulk List-ID: X-Mailing-List: linux-acpi@vger.kernel.org On Tue, Sep 28, 2021 at 07:26:52PM +0200, Rafael J. Wysocki wrote: > On 9/24/2021 11:04 AM, Lorenzo Pieralisi wrote: > > On Thu, Sep 23, 2021 at 02:54:52PM +0200, Rafael J. Wysocki wrote: > > > On Thu, Sep 23, 2021 at 2:26 PM Mark Kettenis wrote: > > > > > From: "Rafael J. Wysocki" > > > > > Date: Thu, 23 Sep 2021 13:05:05 +0200 > > > > > > > > > > On Thu, Sep 23, 2021 at 11:40 AM Lorenzo Pieralisi > > > > > wrote: > > > > > > On Thu, Sep 23, 2021 at 01:09:58AM +0200, Mark Kettenis wrote: > > > > > > > > Date: Wed, 22 Sep 2021 17:33:36 +0100 > > > > > > > > From: Lorenzo Pieralisi > > > > > > > > > > > > > > > > On Fri, Sep 10, 2021 at 10:32:23PM +0800, Jia He wrote: > > > > > > > > > This reverts commit 437b38c51162f8b87beb28a833c4d5dc85fa864e. > > > > > > > > > > > > > > > > > > After this commit, a boot panic is alway hit on an Ampere EMAG server > > > > > > > > > with call trace as follows: > > > > > > > > > Internal error: synchronous external abort: 96000410 [#1] SMP > > > > > > > > > Modules linked in: > > > > > > > > > CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.14.0+ #462 > > > > > > > > > Hardware name: MiTAC RAPTOR EV-883832-X3-0001/RAPTOR, BIOS 0.14 02/22/2019 > > > > > > > > > pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) > > > > > > > > > [...snip...] > > > > > > > > > Call trace: > > > > > > > > > acpi_ex_system_memory_space_handler+0x26c/0x2c8 > > > > > > > > > acpi_ev_address_space_dispatch+0x228/0x2c4 > > > > > > > > > acpi_ex_access_region+0x114/0x268 > > > > > > > > > acpi_ex_field_datum_io+0x128/0x1b8 > > > > > > > > > acpi_ex_extract_from_field+0x14c/0x2ac > > > > > > > > > acpi_ex_read_data_from_field+0x190/0x1b8 > > > > > > > > > acpi_ex_resolve_node_to_value+0x1ec/0x288 > > > > > > > > > acpi_ex_resolve_to_value+0x250/0x274 > > > > > > > > > acpi_ds_evaluate_name_path+0xac/0x124 > > > > > > > > > acpi_ds_exec_end_op+0x90/0x410 > > > > > > > > > acpi_ps_parse_loop+0x4ac/0x5d8 > > > > > > > > > acpi_ps_parse_aml+0xe0/0x2c8 > > > > > > > > > acpi_ps_execute_method+0x19c/0x1ac > > > > > > > > > acpi_ns_evaluate+0x1f8/0x26c > > > > > > > > > acpi_ns_init_one_device+0x104/0x140 > > > > > > > > > acpi_ns_walk_namespace+0x158/0x1d0 > > > > > > > > > acpi_ns_initialize_devices+0x194/0x218 > > > > > > > > > acpi_initialize_objects+0x48/0x50 > > > > > > > > > acpi_init+0xe0/0x498 > > > > > > > > > > > > > > > > > > As mentioned by Lorenzo: > > > > > > > > > "We are forcing memory semantics mappings to PROT_NORMAL_NC, which > > > > > > > > > eMAG does not like at all and I'd need to understand why. It looks > > > > > > > > > like the issue happen in SystemMemory Opregion handler." > > > > > > > > > > > > > > > > > > Hence just revert it before everything is clear. > > > > > > > > > > > > > > > > > > Fixes: 437b38c51162 ("ACPI: Add memory semantics to acpi_os_map_memory()") > > > > > > > > > Cc: Lorenzo Pieralisi > > > > > > > > > Cc: Ard Biesheuvel > > > > > > > > > Cc: Hanjun Guo > > > > > > > > > Cc: Catalin Marinas > > > > > > > > > Cc: Rafael J. Wysocki > > > > > > > > > Cc: Harb Abdulhamid > > > > > > > > > > > > > > > > > > Signed-off-by: Jia He > > > > > > > > Rewrote the commit log, please take the patch below and repost > > > > > > > > it as a v3. > > > > > > > > > > > > > > > > It would still be great if Ampere can help us understand why > > > > > > > > the NormalNC attributes trigger a sync abort on the opregion > > > > > > > > before merging it. > > > > > > > To be honest, I don't think you really need an explanation from Ampere > > > > > > > here. Mapping a part of the address space that doesn't provide memory > > > > > > > semantics with NormalNC attributes is wrong and triggering a sync > > > > > > > abort in that case is way better than silently ignoring the access. > > > > > > That's understood and that's what I explained in the revert commit > > > > > > log, no question about it. > > > > > > > > > > > > I was just asking to confirm if that's what's actually happening. > > > > > > > > > > > > > Putting my OpenBSD hat on (where we have our own ACPI OSPM > > > > > > > implementation) I must say that we always interpreted SystemMemory as > > > > > > > memory mapped IO and I think that is a logical choice as SystemIO is > > > > > > > used for (non-memory mapped) IO. And I'd say that the ACPI OSPM code > > > > > > > should make sure that it uses properly aligned access to any Field > > > > > > > object that doesn't use AnyAcc as its access type. Even on x86! And > > > > > > > I'd say that AML that uses AnyAcc fields for SystemMemory OpRegions on > > > > > > > arm64 is buggy. > > > > > > > > > > > > > > But maybe relaxing this when the EFI memory map indicates that the > > > > > > > address space in question does provide memory semantics does make > > > > > > > sense. That should defenitely be documented in the ACPI standard > > > > > > > though. > > > > > > Mapping SystemMemory Opregions as "memory" does not make sense > > > > > > at all to me. Still, that's what Linux ACPICA code does (*if* > > > > > > that's what acpi_os_map_memory() is supposed to mean). > > > > > > > > > > > > https://lore.kernel.org/linux-acpi/20210916160827.GA4525@lpieralisi > > > > > It doesn't need to do that, though, if there are good enough arguments > > > > > to change the current behavior (and the argument here is that it may > > > > > be an MMIO region, so mapping it as memory doesn't really work, but it > > > > > also may be a region in memory - there is no rule in the spec by which > > > > > SystemMemory Opregions cannot be "memory" AFAICS) and if that change > > > > > doesn't introduce regressions in the installed base. > > > > > > > > > > > Where do we go from here, to be defined, we still have a bug > > > > > > to fix after the revert is applied. > > > > > > > > > > > > drivers/acpi/sysfs.c > > > > > > > > > > > > maps BERT error regions with acpi_os_map_memory(). > > > > > That mechanism is basically used for exporting ACPI tables to user > > > > > space and they are known to reside in memory. Whether or not BERT > > > > > regions should be mapped in the same way is a good question. > > > > It is not inconceivable that BERT regions actually live in memory of > > > > the BMC that is exposed over a bus that doesn't implement memory > > > > semantics is it? > > > No, it isn't, which is why I think that mapping them as RAM may not be > > > a good idea in general. > > Should I patch acpi_data_show() to map BERT error regions (well, that's > > what acpi_data_show() is used on at the moment) as MMIO and use the > > related memcpy routine to read them then :) ? > > It actually would be good to clean it up so it is clear that this is > only used for BERT. I could, I wonder what's best to do that though. Maybe making acpi_table_data_init() acpi_table_bert_data_init() and remove the infrastructure built on top of acpi_data_obj ? I wonder whether adding a bin_attribute.read() pointer in the acpi_data_obj struct (that would make it table specific) would be the most elegant solution (even though the whole infrastructure has been used only for BERT for quite a while). Lorenzo 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 D0133C433FE for ; Tue, 2 Nov 2021 15:12:46 +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 A1ED560F02 for ; Tue, 2 Nov 2021 15:12:46 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.1 mail.kernel.org A1ED560F02 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=arm.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=bLszT1HoVCxdvOdVXbvyob6ZrSPntl1ebwzf77h8JXU=; b=LIQyosJ3KQ34az kSbNku+tzQCvLo92ci+hI2SI0S5CXtkmXOcr2ORm/PI2HGA7+Gez5fLk8cOkpl8ekEKNx5hPjHmZt TzafrkKsGS16EMQrxNy9+hPzprMXpH5ZpNOh5DXorZSbIsBXqDLensLTIC8Q2n8NJWXKZsm1X1CxR PEIKtyw8FkhOakUMM2kgOZZHtiFEV/9QpISBOg2uocSPq3jzJNxi9DJyX4npOFUELYQOKxjP0/LWy fGFNictD7v6oe/ZgYNs1jfimZiuEqepj856IpKD4CVhY1HNiOknXrrklMcWynn81zDTx2PcU3ofFC EIzr3BUDXW8XqRas2Eug==; Received: from localhost ([::1] helo=bombadil.infradead.org) by bombadil.infradead.org with esmtp (Exim 4.94.2 #2 (Red Hat Linux)) id 1mhvRf-0025Ws-GO; Tue, 02 Nov 2021 15:11:27 +0000 Received: from foss.arm.com ([217.140.110.172]) by bombadil.infradead.org with esmtp (Exim 4.94.2 #2 (Red Hat Linux)) id 1mhvRa-0025VC-B7 for linux-arm-kernel@lists.infradead.org; Tue, 02 Nov 2021 15:11:24 +0000 Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.121.207.14]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id C557DD6E; Tue, 2 Nov 2021 08:11:18 -0700 (PDT) Received: from lpieralisi (e121166-lin.cambridge.arm.com [10.1.196.255]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id C913B3F7B4; Tue, 2 Nov 2021 08:11:16 -0700 (PDT) Date: Tue, 2 Nov 2021 15:11:14 +0000 From: Lorenzo Pieralisi To: "Rafael J. Wysocki" Cc: "Rafael J. Wysocki" , Mark Kettenis , Jia He , Harb Abdulhamid , Will Deacon , Len Brown , Robert Moore , Erik Kaneda , Linux ARM , Linux Kernel Mailing List , ACPI Devel Maling List , "open list:ACPI COMPONENT ARCHITECTURE (ACPICA)" , Ard Biesheuvel , Hanjun Guo , Catalin Marinas Subject: Re: [PATCH v2] Revert "ACPI: Add memory semantics to acpi_os_map_memory()" Message-ID: <20211102151114.GB15962@lpieralisi> References: <20210910122820.26886-1-justin.he@arm.com> <20210910143223.6705-1-justin.he@arm.com> <20210922163336.GA24633@lpieralisi> <56147a0b8b9fba46@bloch.sibelius.xs4all.nl> <20210923094031.GA6454@lpieralisi> <56147c6e73afe9f6@bloch.sibelius.xs4all.nl> <20210924090409.GA26089@lpieralisi> <44f20f07-257b-a1a5-23d9-ffd66bf45887@intel.com> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <44f20f07-257b-a1a5-23d9-ffd66bf45887@intel.com> User-Agent: Mutt/1.9.4 (2018-02-28) X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20211102_081122_528782_E9CDEAD5 X-CRM114-Status: GOOD ( 55.99 ) 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 Tue, Sep 28, 2021 at 07:26:52PM +0200, Rafael J. Wysocki wrote: > On 9/24/2021 11:04 AM, Lorenzo Pieralisi wrote: > > On Thu, Sep 23, 2021 at 02:54:52PM +0200, Rafael J. Wysocki wrote: > > > On Thu, Sep 23, 2021 at 2:26 PM Mark Kettenis wrote: > > > > > From: "Rafael J. Wysocki" > > > > > Date: Thu, 23 Sep 2021 13:05:05 +0200 > > > > > > > > > > On Thu, Sep 23, 2021 at 11:40 AM Lorenzo Pieralisi > > > > > wrote: > > > > > > On Thu, Sep 23, 2021 at 01:09:58AM +0200, Mark Kettenis wrote: > > > > > > > > Date: Wed, 22 Sep 2021 17:33:36 +0100 > > > > > > > > From: Lorenzo Pieralisi > > > > > > > > > > > > > > > > On Fri, Sep 10, 2021 at 10:32:23PM +0800, Jia He wrote: > > > > > > > > > This reverts commit 437b38c51162f8b87beb28a833c4d5dc85fa864e. > > > > > > > > > > > > > > > > > > After this commit, a boot panic is alway hit on an Ampere EMAG server > > > > > > > > > with call trace as follows: > > > > > > > > > Internal error: synchronous external abort: 96000410 [#1] SMP > > > > > > > > > Modules linked in: > > > > > > > > > CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.14.0+ #462 > > > > > > > > > Hardware name: MiTAC RAPTOR EV-883832-X3-0001/RAPTOR, BIOS 0.14 02/22/2019 > > > > > > > > > pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) > > > > > > > > > [...snip...] > > > > > > > > > Call trace: > > > > > > > > > acpi_ex_system_memory_space_handler+0x26c/0x2c8 > > > > > > > > > acpi_ev_address_space_dispatch+0x228/0x2c4 > > > > > > > > > acpi_ex_access_region+0x114/0x268 > > > > > > > > > acpi_ex_field_datum_io+0x128/0x1b8 > > > > > > > > > acpi_ex_extract_from_field+0x14c/0x2ac > > > > > > > > > acpi_ex_read_data_from_field+0x190/0x1b8 > > > > > > > > > acpi_ex_resolve_node_to_value+0x1ec/0x288 > > > > > > > > > acpi_ex_resolve_to_value+0x250/0x274 > > > > > > > > > acpi_ds_evaluate_name_path+0xac/0x124 > > > > > > > > > acpi_ds_exec_end_op+0x90/0x410 > > > > > > > > > acpi_ps_parse_loop+0x4ac/0x5d8 > > > > > > > > > acpi_ps_parse_aml+0xe0/0x2c8 > > > > > > > > > acpi_ps_execute_method+0x19c/0x1ac > > > > > > > > > acpi_ns_evaluate+0x1f8/0x26c > > > > > > > > > acpi_ns_init_one_device+0x104/0x140 > > > > > > > > > acpi_ns_walk_namespace+0x158/0x1d0 > > > > > > > > > acpi_ns_initialize_devices+0x194/0x218 > > > > > > > > > acpi_initialize_objects+0x48/0x50 > > > > > > > > > acpi_init+0xe0/0x498 > > > > > > > > > > > > > > > > > > As mentioned by Lorenzo: > > > > > > > > > "We are forcing memory semantics mappings to PROT_NORMAL_NC, which > > > > > > > > > eMAG does not like at all and I'd need to understand why. It looks > > > > > > > > > like the issue happen in SystemMemory Opregion handler." > > > > > > > > > > > > > > > > > > Hence just revert it before everything is clear. > > > > > > > > > > > > > > > > > > Fixes: 437b38c51162 ("ACPI: Add memory semantics to acpi_os_map_memory()") > > > > > > > > > Cc: Lorenzo Pieralisi > > > > > > > > > Cc: Ard Biesheuvel > > > > > > > > > Cc: Hanjun Guo > > > > > > > > > Cc: Catalin Marinas > > > > > > > > > Cc: Rafael J. Wysocki > > > > > > > > > Cc: Harb Abdulhamid > > > > > > > > > > > > > > > > > > Signed-off-by: Jia He > > > > > > > > Rewrote the commit log, please take the patch below and repost > > > > > > > > it as a v3. > > > > > > > > > > > > > > > > It would still be great if Ampere can help us understand why > > > > > > > > the NormalNC attributes trigger a sync abort on the opregion > > > > > > > > before merging it. > > > > > > > To be honest, I don't think you really need an explanation from Ampere > > > > > > > here. Mapping a part of the address space that doesn't provide memory > > > > > > > semantics with NormalNC attributes is wrong and triggering a sync > > > > > > > abort in that case is way better than silently ignoring the access. > > > > > > That's understood and that's what I explained in the revert commit > > > > > > log, no question about it. > > > > > > > > > > > > I was just asking to confirm if that's what's actually happening. > > > > > > > > > > > > > Putting my OpenBSD hat on (where we have our own ACPI OSPM > > > > > > > implementation) I must say that we always interpreted SystemMemory as > > > > > > > memory mapped IO and I think that is a logical choice as SystemIO is > > > > > > > used for (non-memory mapped) IO. And I'd say that the ACPI OSPM code > > > > > > > should make sure that it uses properly aligned access to any Field > > > > > > > object that doesn't use AnyAcc as its access type. Even on x86! And > > > > > > > I'd say that AML that uses AnyAcc fields for SystemMemory OpRegions on > > > > > > > arm64 is buggy. > > > > > > > > > > > > > > But maybe relaxing this when the EFI memory map indicates that the > > > > > > > address space in question does provide memory semantics does make > > > > > > > sense. That should defenitely be documented in the ACPI standard > > > > > > > though. > > > > > > Mapping SystemMemory Opregions as "memory" does not make sense > > > > > > at all to me. Still, that's what Linux ACPICA code does (*if* > > > > > > that's what acpi_os_map_memory() is supposed to mean). > > > > > > > > > > > > https://lore.kernel.org/linux-acpi/20210916160827.GA4525@lpieralisi > > > > > It doesn't need to do that, though, if there are good enough arguments > > > > > to change the current behavior (and the argument here is that it may > > > > > be an MMIO region, so mapping it as memory doesn't really work, but it > > > > > also may be a region in memory - there is no rule in the spec by which > > > > > SystemMemory Opregions cannot be "memory" AFAICS) and if that change > > > > > doesn't introduce regressions in the installed base. > > > > > > > > > > > Where do we go from here, to be defined, we still have a bug > > > > > > to fix after the revert is applied. > > > > > > > > > > > > drivers/acpi/sysfs.c > > > > > > > > > > > > maps BERT error regions with acpi_os_map_memory(). > > > > > That mechanism is basically used for exporting ACPI tables to user > > > > > space and they are known to reside in memory. Whether or not BERT > > > > > regions should be mapped in the same way is a good question. > > > > It is not inconceivable that BERT regions actually live in memory of > > > > the BMC that is exposed over a bus that doesn't implement memory > > > > semantics is it? > > > No, it isn't, which is why I think that mapping them as RAM may not be > > > a good idea in general. > > Should I patch acpi_data_show() to map BERT error regions (well, that's > > what acpi_data_show() is used on at the moment) as MMIO and use the > > related memcpy routine to read them then :) ? > > It actually would be good to clean it up so it is clear that this is > only used for BERT. I could, I wonder what's best to do that though. Maybe making acpi_table_data_init() acpi_table_bert_data_init() and remove the infrastructure built on top of acpi_data_obj ? I wonder whether adding a bin_attribute.read() pointer in the acpi_data_obj struct (that would make it table specific) would be the most elegant solution (even though the whole infrastructure has been used only for BERT for quite a while). Lorenzo _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============5215695862068938242==" MIME-Version: 1.0 From: Lorenzo Pieralisi Subject: [Devel] Re: [PATCH v2] Revert "ACPI: Add memory semantics to acpi_os_map_memory()" Date: Tue, 02 Nov 2021 15:11:14 +0000 Message-ID: <20211102151114.GB15962@lpieralisi> In-Reply-To: 44f20f07-257b-a1a5-23d9-ffd66bf45887@intel.com List-ID: To: devel@acpica.org --===============5215695862068938242== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On Tue, Sep 28, 2021 at 07:26:52PM +0200, Rafael J. Wysocki wrote: > On 9/24/2021 11:04 AM, Lorenzo Pieralisi wrote: > > On Thu, Sep 23, 2021 at 02:54:52PM +0200, Rafael J. Wysocki wrote: > > > On Thu, Sep 23, 2021 at 2:26 PM Mark Kettenis wrote: > > > > > From: "Rafael J. Wysocki" > > > > > Date: Thu, 23 Sep 2021 13:05:05 +0200 > > > > > = > > > > > On Thu, Sep 23, 2021 at 11:40 AM Lorenzo Pieralisi > > > > > wrote: > > > > > > On Thu, Sep 23, 2021 at 01:09:58AM +0200, Mark Kettenis wrote: > > > > > > > > Date: Wed, 22 Sep 2021 17:33:36 +0100 > > > > > > > > From: Lorenzo Pieralisi > > > > > > > > = > > > > > > > > On Fri, Sep 10, 2021 at 10:32:23PM +0800, Jia He wrote: > > > > > > > > > This reverts commit 437b38c51162f8b87beb28a833c4d5dc85fa8= 64e. > > > > > > > > > = > > > > > > > > > After this commit, a boot panic is alway hit on an Ampere= EMAG server > > > > > > > > > with call trace as follows: > > > > > > > > > Internal error: synchronous external abort: 96000410 [#= 1] SMP > > > > > > > > > Modules linked in: > > > > > > > > > CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.14.0+ #462 > > > > > > > > > Hardware name: MiTAC RAPTOR EV-883832-X3-0001/RAPTOR, B= IOS 0.14 02/22/2019 > > > > > > > > > pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS B= TYPE=3D--) > > > > > > > > > [...snip...] > > > > > > > > > Call trace: > > > > > > > > > acpi_ex_system_memory_space_handler+0x26c/0x2c8 > > > > > > > > > acpi_ev_address_space_dispatch+0x228/0x2c4 > > > > > > > > > acpi_ex_access_region+0x114/0x268 > > > > > > > > > acpi_ex_field_datum_io+0x128/0x1b8 > > > > > > > > > acpi_ex_extract_from_field+0x14c/0x2ac > > > > > > > > > acpi_ex_read_data_from_field+0x190/0x1b8 > > > > > > > > > acpi_ex_resolve_node_to_value+0x1ec/0x288 > > > > > > > > > acpi_ex_resolve_to_value+0x250/0x274 > > > > > > > > > acpi_ds_evaluate_name_path+0xac/0x124 > > > > > > > > > acpi_ds_exec_end_op+0x90/0x410 > > > > > > > > > acpi_ps_parse_loop+0x4ac/0x5d8 > > > > > > > > > acpi_ps_parse_aml+0xe0/0x2c8 > > > > > > > > > acpi_ps_execute_method+0x19c/0x1ac > > > > > > > > > acpi_ns_evaluate+0x1f8/0x26c > > > > > > > > > acpi_ns_init_one_device+0x104/0x140 > > > > > > > > > acpi_ns_walk_namespace+0x158/0x1d0 > > > > > > > > > acpi_ns_initialize_devices+0x194/0x218 > > > > > > > > > acpi_initialize_objects+0x48/0x50 > > > > > > > > > acpi_init+0xe0/0x498 > > > > > > > > > = > > > > > > > > > As mentioned by Lorenzo: > > > > > > > > > "We are forcing memory semantics mappings to PROT_NORM= AL_NC, which > > > > > > > > > eMAG does not like at all and I'd need to understand w= hy. It looks > > > > > > > > > like the issue happen in SystemMemory Opregion handler= ." > > > > > > > > > = > > > > > > > > > Hence just revert it before everything is clear. > > > > > > > > > = > > > > > > > > > Fixes: 437b38c51162 ("ACPI: Add memory semantics to acpi_= os_map_memory()") > > > > > > > > > Cc: Lorenzo Pieralisi > > > > > > > > > Cc: Ard Biesheuvel > > > > > > > > > Cc: Hanjun Guo > > > > > > > > > Cc: Catalin Marinas > > > > > > > > > Cc: Rafael J. Wysocki > > > > > > > > > Cc: Harb Abdulhamid > > > > > > > > > = > > > > > > > > > Signed-off-by: Jia He > > > > > > > > Rewrote the commit log, please take the patch below and rep= ost > > > > > > > > it as a v3. > > > > > > > > = > > > > > > > > It would still be great if Ampere can help us understand why > > > > > > > > the NormalNC attributes trigger a sync abort on the opregion > > > > > > > > before merging it. > > > > > > > To be honest, I don't think you really need an explanation fr= om Ampere > > > > > > > here. Mapping a part of the address space that doesn't provi= de memory > > > > > > > semantics with NormalNC attributes is wrong and triggering a = sync > > > > > > > abort in that case is way better than silently ignoring the a= ccess. > > > > > > That's understood and that's what I explained in the revert com= mit > > > > > > log, no question about it. > > > > > > = > > > > > > I was just asking to confirm if that's what's actually happenin= g. > > > > > > = > > > > > > > Putting my OpenBSD hat on (where we have our own ACPI OSPM > > > > > > > implementation) I must say that we always interpreted SystemM= emory as > > > > > > > memory mapped IO and I think that is a logical choice as Syst= emIO is > > > > > > > used for (non-memory mapped) IO. And I'd say that the ACPI O= SPM code > > > > > > > should make sure that it uses properly aligned access to any = Field > > > > > > > object that doesn't use AnyAcc as its access type. Even on x= 86! And > > > > > > > I'd say that AML that uses AnyAcc fields for SystemMemory OpR= egions on > > > > > > > arm64 is buggy. > > > > > > > = > > > > > > > But maybe relaxing this when the EFI memory map indicates tha= t the > > > > > > > address space in question does provide memory semantics does = make > > > > > > > sense. That should defenitely be documented in the ACPI stan= dard > > > > > > > though. > > > > > > Mapping SystemMemory Opregions as "memory" does not make sense > > > > > > at all to me. Still, that's what Linux ACPICA code does (*if* > > > > > > that's what acpi_os_map_memory() is supposed to mean). > > > > > > = > > > > > > https://lore.kernel.org/linux-acpi/20210916160827.GA4525(a)lpie= ralisi > > > > > It doesn't need to do that, though, if there are good enough argu= ments > > > > > to change the current behavior (and the argument here is that it = may > > > > > be an MMIO region, so mapping it as memory doesn't really work, b= ut it > > > > > also may be a region in memory - there is no rule in the spec by = which > > > > > SystemMemory Opregions cannot be "memory" AFAICS) and if that cha= nge > > > > > doesn't introduce regressions in the installed base. > > > > > = > > > > > > Where do we go from here, to be defined, we still have a bug > > > > > > to fix after the revert is applied. > > > > > > = > > > > > > drivers/acpi/sysfs.c > > > > > > = > > > > > > maps BERT error regions with acpi_os_map_memory(). > > > > > That mechanism is basically used for exporting ACPI tables to user > > > > > space and they are known to reside in memory. Whether or not BERT > > > > > regions should be mapped in the same way is a good question. > > > > It is not inconceivable that BERT regions actually live in memory of > > > > the BMC that is exposed over a bus that doesn't implement memory > > > > semantics is it? > > > No, it isn't, which is why I think that mapping them as RAM may not be > > > a good idea in general. > > Should I patch acpi_data_show() to map BERT error regions (well, that's > > what acpi_data_show() is used on at the moment) as MMIO and use the > > related memcpy routine to read them then :) ? > = > It actually would be good to clean it up so it is clear that this is > only used for BERT. I could, I wonder what's best to do that though. Maybe making acpi_table_data_init() acpi_table_bert_data_init() and remove the infrastructure built on top of acpi_data_obj ? I wonder whether adding a bin_attribute.read() pointer in the acpi_data_obj struct (that would make it table specific) would be the most elegant solution (even though the whole infrastructure has been used only for BERT for quite a while). Lorenzo --===============5215695862068938242==--