Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751601AbdHQIxA (ORCPT ); Thu, 17 Aug 2017 04:53:00 -0400 Received: from usa-sjc-mx-foss1.foss.arm.com ([217.140.101.70]:46180 "EHLO foss.arm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750939AbdHQIw6 (ORCPT ); Thu, 17 Aug 2017 04:52:58 -0400 Subject: Re: [PATCH] irqchip/gic-{v2m, v3-its}: check iommu capable before doing iommu map To: Shawn Lin , Thomas Gleixner Cc: Jason Cooper , linux-kernel@vger.kernel.org, Joerg Roedel , Robin Murphy , iommu@lists.linux-foundation.org References: <1502958508-195670-1-git-send-email-shawn.lin@rock-chips.com> From: Marc Zyngier Organization: ARM Ltd Message-ID: <8017a14e-0855-de2c-9a78-5abeab25aeb1@arm.com> Date: Thu, 17 Aug 2017 09:52:55 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.2.1 MIME-Version: 1.0 In-Reply-To: <1502958508-195670-1-git-send-email-shawn.lin@rock-chips.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-GB Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 3446 Lines: 78 On 17/08/17 09:28, Shawn Lin wrote: > If a PCIe RC use gic-v2m or gic-v3-its as a msi domain but doesn't > have iommu support, we don't need to do iommu_dma_map_msi_msg to > get mapped iommu address as all we need is the physical address. > Otherwise we see the oops shown below as we can't get a iommu_group > for a device without iommu capable. > > Unable to handle kernel NULL pointer dereference at virtual address > 000000d0 > [00000000000000d0] user address but active_mm is swapper > Internal error: Oops: 96000004 [#1] PREEMPT SMP > Modules linked in: > CPU: 3 PID: 1 Comm: swapper/0 Not tainted > 4.13.0-rc5-next-20170815-00001-g0744890-dirty #53 > Hardware name: Firefly-RK3399 Board (DT) > task: ffff80007bc70000 task.stack: ffff80007bc6c000 > PC is at iommu_get_domain_for_dev+0x38/0x50 > LR is at iommu_dma_map_msi_msg+0x3c/0x1b8 > pc : [] lr : [] pstate: > a0000045 > > ... > > [] iommu_get_domain_for_dev+0x38/0x50 > [] iommu_dma_map_msi_msg+0x3c/0x1b8 > [] its_irq_compose_msi_msg+0x44/0x50 > [] irq_chip_compose_msi_msg+0x40/0x58 > [] msi_domain_activate+0x1c/0x48 > [] __irq_domain_activate_irq+0x40/0x58 > [] irq_domain_activate_irq+0x24/0x40 > [] msi_domain_alloc_irqs+0x104/0x190 > [] pci_msi_setup_msi_irqs+0x3c/0x48 > [] __pci_enable_msi_range+0x21c/0x408 > [] pci_alloc_irq_vectors_affinity+0x104/0x168 > [] pcie_port_device_register+0x200/0x488 > [] pcie_portdrv_probe+0x34/0xd0 > [] local_pci_probe+0x3c/0xb8 > [] pci_device_probe+0x138/0x170 > [] driver_probe_device+0x21c/0x2d8 > [] __device_attach_driver+0x9c/0xf8 > [] bus_for_each_drv+0x5c/0x98 > [] __device_attach+0xc4/0x138 > [] device_attach+0x10/0x18 > [] pci_bus_add_device+0x4c/0xa8 > [] pci_bus_add_devices+0x44/0x90 > [] rockchip_pcie_probe+0xc70/0xcc8 > [] platform_drv_probe+0x58/0xc0 > [] driver_probe_device+0x21c/0x2d8 > [] __driver_attach+0xac/0xb0 > [] bus_for_each_dev+0x64/0xa0 > [] driver_attach+0x20/0x28 > [] bus_add_driver+0x110/0x230 > [] driver_register+0x60/0xf8 > [] __platform_driver_register+0x40/0x48 > [] rockchip_pcie_driver_init+0x18/0x20 > [] do_one_initcall+0xb0/0x120 > [] kernel_init_freeable+0x184/0x224 > [] kernel_init+0x10/0x100 > [] ret_from_fork+0x10/0x18 > > This patch is to fix the problem exposed by the commit mentioned below. > Before this commit, iommu has a work around method to fix this but now > it doesn't. So we could fix this in gic code but maybe still need a fixes > tag here. > > Fixes: 05f80300dc8b ("iommu: Finish making iommu_group support mandatory") > Signed-off-by: Shawn Lin That looks pretty horrible. Please see the patch I posted in response to your initial report: http://marc.info/?l=linux-pci&m=150295809430903&w=2 Thanks, M. -- Jazz is not dead. It just smells funny...