From: Lu Baolu <baolu.lu@linux.intel.com> To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>, Joerg Roedel <joro@8bytes.org>, Alex Williamson <alex.williamson@redhat.com>, Bjorn Helgaas <bhelgaas@google.com>, Jason Gunthorpe <jgg@nvidia.com>, Kevin Tian <kevin.tian@intel.com>, Ashok Raj <ashok.raj@intel.com> Cc: Will Deacon <will@kernel.org>, rafael@kernel.org, Diana Craciun <diana.craciun@oss.nxp.com>, Cornelia Huck <cohuck@redhat.com>, Eric Auger <eric.auger@redhat.com>, Liu Yi L <yi.l.liu@intel.com>, Jacob jun Pan <jacob.jun.pan@intel.com>, Chaitanya Kulkarni <kch@nvidia.com>, iommu@lists.linux-foundation.org, linux-pci@vger.kernel.org, kvm@vger.kernel.org, linux-kernel@vger.kernel.org, Lu Baolu <baolu.lu@linux.intel.com> Subject: [PATCH 00/11] Fix BUG_ON in vfio_iommu_group_notifier() Date: Mon, 15 Nov 2021 10:05:41 +0800 [thread overview] Message-ID: <20211115020552.2378167-1-baolu.lu@linux.intel.com> (raw) Hi folks, The iommu group is the minimal isolation boundary for DMA. Devices in a group can access each other's MMIO registers via peer to peer DMA and also need share the same I/O address space. Once the I/O address space is assigned to user control it is no longer available to the dma_map* API, which effectively makes the DMA API non-working. Second, userspace can use DMA initiated by a device that it controls to access the MMIO spaces of other devices in the group. This allows userspace to indirectly attack any kernel owned device and it's driver. Therefore groups must either be entirely under kernel control or userspace control, never a mixture. Unfortunately some systems have problems with the granularity of groups and there are a couple of important exceptions: - pci_stub allows the admin to block driver binding on a device and make it permanently shared with userspace. Since PCI stub does not do DMA it is safe, however the admin must understand that using pci_stub allows userspace to attack whatever device it was bound it. - PCI bridges are sometimes included in groups. Typically PCI bridges do not use DMA, and generally do not have MMIO regions. Generally any device that does not have any MMIO registers is a possible candidate for an exception. Currently vfio adopts a workaround to detect violations of the above restrictions by monitoring the driver core BOUND event, and hardwiring the above exceptions. Since there is no way for vfio to reject driver binding at this point, BUG_ON() is triggered if a violation is captured (kernel driver BOUND event on a group which already has some devices assigned to userspace). Aside from the bad user experience this opens a way for root userspace to crash the kernel, even in high integrity configurations, by manipulating the module binding and triggering the BUG_ON. This series solves this problem by making the user/kernel ownership a core concept at the IOMMU layer. The driver core enforces kernel ownership while drivers are bound and violations now result in a error codes during probe, not BUG_ON failures. Patch partitions: [PATCH 1-2]: Detect DMA ownership conflicts during driver binding; [PATCH 3-6]: Add security context management for assigned devices; [PATCH 7-11]: Various cleanups. Ideas contributed by: Jason Gunthorpe <jgg@nvidia.com> Kevin Tian <kevin.tian@intel.com> Ashok Raj <ashok.raj@intel.com> Lu Baolu <baolu.lu@linux.intel.com> Review contributors: Jason Gunthorpe <jgg@nvidia.com> Kevin Tian <kevin.tian@intel.com> Ashok Raj <ashok.raj@intel.com> Liu Yi L <yi.l.liu@intel.com> Jacob jun Pan <jacob.jun.pan@intel.com> Chaitanya Kulkarni <kch@nvidia.com> This also is part one of three initial series for IOMMUFD: * Move IOMMU Group security into the iommu layer - Generic IOMMUFD implementation - VFIO ability to consume IOMMUFD This is based on v5.16-rc1 and available on github: https://github.com/LuBaolu/intel-iommu/commits/iommu-dma-ownership-v1 Best regards, baolu Jason Gunthorpe (1): vfio: Delete the unbound_list Lu Baolu (10): iommu: Add device dma ownership set/release interfaces driver core: Set DMA ownership during driver bind/unbind PCI: pci_stub: Suppress kernel DMA ownership auto-claiming PCI: portdrv: Suppress kernel DMA ownership auto-claiming iommu: Add security context management for assigned devices iommu: Expose group variants of dma ownership interfaces vfio: Use DMA_OWNER_USER to declaim passthrough devices vfio: Remove use of vfio_group_viable() vfio: Remove iommu group notifier iommu: Remove iommu group changes notifier include/linux/device/driver.h | 7 +- include/linux/iommu.h | 75 ++++--- drivers/base/dd.c | 12 ++ drivers/iommu/iommu.c | 274 ++++++++++++++++++-------- drivers/pci/pci-stub.c | 3 + drivers/pci/pcie/portdrv_pci.c | 2 + drivers/vfio/fsl-mc/vfio_fsl_mc.c | 1 + drivers/vfio/pci/vfio_pci.c | 3 + drivers/vfio/platform/vfio_amba.c | 1 + drivers/vfio/platform/vfio_platform.c | 1 + drivers/vfio/vfio.c | 247 ++--------------------- 11 files changed, 294 insertions(+), 332 deletions(-) -- 2.25.1
WARNING: multiple messages have this Message-ID (diff)
From: Lu Baolu <baolu.lu@linux.intel.com> To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>, Joerg Roedel <joro@8bytes.org>, Alex Williamson <alex.williamson@redhat.com>, Bjorn Helgaas <bhelgaas@google.com>, Jason Gunthorpe <jgg@nvidia.com>, Kevin Tian <kevin.tian@intel.com>, Ashok Raj <ashok.raj@intel.com> Cc: Chaitanya Kulkarni <kch@nvidia.com>, kvm@vger.kernel.org, rafael@kernel.org, linux-pci@vger.kernel.org, Cornelia Huck <cohuck@redhat.com>, linux-kernel@vger.kernel.org, iommu@lists.linux-foundation.org, Jacob jun Pan <jacob.jun.pan@intel.com>, Diana Craciun <diana.craciun@oss.nxp.com>, Will Deacon <will@kernel.org> Subject: [PATCH 00/11] Fix BUG_ON in vfio_iommu_group_notifier() Date: Mon, 15 Nov 2021 10:05:41 +0800 [thread overview] Message-ID: <20211115020552.2378167-1-baolu.lu@linux.intel.com> (raw) Hi folks, The iommu group is the minimal isolation boundary for DMA. Devices in a group can access each other's MMIO registers via peer to peer DMA and also need share the same I/O address space. Once the I/O address space is assigned to user control it is no longer available to the dma_map* API, which effectively makes the DMA API non-working. Second, userspace can use DMA initiated by a device that it controls to access the MMIO spaces of other devices in the group. This allows userspace to indirectly attack any kernel owned device and it's driver. Therefore groups must either be entirely under kernel control or userspace control, never a mixture. Unfortunately some systems have problems with the granularity of groups and there are a couple of important exceptions: - pci_stub allows the admin to block driver binding on a device and make it permanently shared with userspace. Since PCI stub does not do DMA it is safe, however the admin must understand that using pci_stub allows userspace to attack whatever device it was bound it. - PCI bridges are sometimes included in groups. Typically PCI bridges do not use DMA, and generally do not have MMIO regions. Generally any device that does not have any MMIO registers is a possible candidate for an exception. Currently vfio adopts a workaround to detect violations of the above restrictions by monitoring the driver core BOUND event, and hardwiring the above exceptions. Since there is no way for vfio to reject driver binding at this point, BUG_ON() is triggered if a violation is captured (kernel driver BOUND event on a group which already has some devices assigned to userspace). Aside from the bad user experience this opens a way for root userspace to crash the kernel, even in high integrity configurations, by manipulating the module binding and triggering the BUG_ON. This series solves this problem by making the user/kernel ownership a core concept at the IOMMU layer. The driver core enforces kernel ownership while drivers are bound and violations now result in a error codes during probe, not BUG_ON failures. Patch partitions: [PATCH 1-2]: Detect DMA ownership conflicts during driver binding; [PATCH 3-6]: Add security context management for assigned devices; [PATCH 7-11]: Various cleanups. Ideas contributed by: Jason Gunthorpe <jgg@nvidia.com> Kevin Tian <kevin.tian@intel.com> Ashok Raj <ashok.raj@intel.com> Lu Baolu <baolu.lu@linux.intel.com> Review contributors: Jason Gunthorpe <jgg@nvidia.com> Kevin Tian <kevin.tian@intel.com> Ashok Raj <ashok.raj@intel.com> Liu Yi L <yi.l.liu@intel.com> Jacob jun Pan <jacob.jun.pan@intel.com> Chaitanya Kulkarni <kch@nvidia.com> This also is part one of three initial series for IOMMUFD: * Move IOMMU Group security into the iommu layer - Generic IOMMUFD implementation - VFIO ability to consume IOMMUFD This is based on v5.16-rc1 and available on github: https://github.com/LuBaolu/intel-iommu/commits/iommu-dma-ownership-v1 Best regards, baolu Jason Gunthorpe (1): vfio: Delete the unbound_list Lu Baolu (10): iommu: Add device dma ownership set/release interfaces driver core: Set DMA ownership during driver bind/unbind PCI: pci_stub: Suppress kernel DMA ownership auto-claiming PCI: portdrv: Suppress kernel DMA ownership auto-claiming iommu: Add security context management for assigned devices iommu: Expose group variants of dma ownership interfaces vfio: Use DMA_OWNER_USER to declaim passthrough devices vfio: Remove use of vfio_group_viable() vfio: Remove iommu group notifier iommu: Remove iommu group changes notifier include/linux/device/driver.h | 7 +- include/linux/iommu.h | 75 ++++--- drivers/base/dd.c | 12 ++ drivers/iommu/iommu.c | 274 ++++++++++++++++++-------- drivers/pci/pci-stub.c | 3 + drivers/pci/pcie/portdrv_pci.c | 2 + drivers/vfio/fsl-mc/vfio_fsl_mc.c | 1 + drivers/vfio/pci/vfio_pci.c | 3 + drivers/vfio/platform/vfio_amba.c | 1 + drivers/vfio/platform/vfio_platform.c | 1 + drivers/vfio/vfio.c | 247 ++--------------------- 11 files changed, 294 insertions(+), 332 deletions(-) -- 2.25.1 _______________________________________________ iommu mailing list iommu@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/iommu
next reply other threads:[~2021-11-15 2:10 UTC|newest] Thread overview: 120+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-11-15 2:05 Lu Baolu [this message] 2021-11-15 2:05 ` [PATCH 00/11] Fix BUG_ON in vfio_iommu_group_notifier() Lu Baolu 2021-11-15 2:05 ` [PATCH 01/11] iommu: Add device dma ownership set/release interfaces Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 13:14 ` Christoph Hellwig 2021-11-15 13:14 ` Christoph Hellwig 2021-11-16 1:57 ` Lu Baolu 2021-11-16 1:57 ` Lu Baolu 2021-11-16 13:46 ` Jason Gunthorpe 2021-11-16 13:46 ` Jason Gunthorpe via iommu 2021-11-17 5:22 ` Lu Baolu 2021-11-17 5:22 ` Lu Baolu 2021-11-17 13:35 ` Jason Gunthorpe 2021-11-17 13:35 ` Jason Gunthorpe via iommu 2021-11-18 1:12 ` Lu Baolu 2021-11-18 1:12 ` Lu Baolu 2021-11-18 14:10 ` Jason Gunthorpe 2021-11-18 14:10 ` Jason Gunthorpe via iommu 2021-11-18 2:39 ` Tian, Kevin 2021-11-18 2:39 ` Tian, Kevin 2021-11-18 13:33 ` Jason Gunthorpe 2021-11-18 13:33 ` Jason Gunthorpe via iommu 2021-11-19 5:44 ` Tian, Kevin 2021-11-19 5:44 ` Tian, Kevin 2021-11-19 11:14 ` Lu Baolu 2021-11-19 11:14 ` Lu Baolu 2021-11-19 15:06 ` Jörg Rödel 2021-11-19 15:06 ` Jörg Rödel 2021-11-19 15:43 ` Jason Gunthorpe 2021-11-19 15:43 ` Jason Gunthorpe via iommu 2021-11-20 11:16 ` Lu Baolu 2021-11-20 11:16 ` Lu Baolu 2021-11-19 12:56 ` Jason Gunthorpe 2021-11-19 12:56 ` Jason Gunthorpe via iommu 2021-11-15 20:38 ` Bjorn Helgaas 2021-11-15 20:38 ` Bjorn Helgaas 2021-11-16 1:52 ` Lu Baolu 2021-11-16 1:52 ` Lu Baolu 2021-11-15 2:05 ` [PATCH 02/11] driver core: Set DMA ownership during driver bind/unbind Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 6:59 ` Greg Kroah-Hartman 2021-11-15 6:59 ` Greg Kroah-Hartman 2021-11-15 13:20 ` Christoph Hellwig 2021-11-15 13:20 ` Christoph Hellwig 2021-11-15 13:38 ` Jason Gunthorpe via iommu 2021-11-15 13:38 ` Jason Gunthorpe 2021-11-15 13:19 ` Christoph Hellwig 2021-11-15 13:19 ` Christoph Hellwig 2021-11-15 13:24 ` Jason Gunthorpe 2021-11-15 13:24 ` Jason Gunthorpe via iommu 2021-11-15 15:37 ` Robin Murphy 2021-11-15 15:37 ` Robin Murphy 2021-11-15 15:56 ` Jason Gunthorpe 2021-11-15 15:56 ` Jason Gunthorpe via iommu 2021-11-15 18:15 ` Christoph Hellwig 2021-11-15 18:15 ` Christoph Hellwig 2021-11-15 18:35 ` Robin Murphy 2021-11-15 18:35 ` Robin Murphy 2021-11-15 19:39 ` Jason Gunthorpe via iommu 2021-11-15 19:39 ` Jason Gunthorpe 2021-11-15 2:05 ` [PATCH 03/11] PCI: pci_stub: Suppress kernel DMA ownership auto-claiming Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 13:21 ` Christoph Hellwig 2021-11-15 13:21 ` Christoph Hellwig 2021-11-15 13:31 ` Jason Gunthorpe via iommu 2021-11-15 13:31 ` Jason Gunthorpe 2021-11-15 15:14 ` Robin Murphy 2021-11-15 15:14 ` Robin Murphy 2021-11-15 16:17 ` Jason Gunthorpe 2021-11-15 16:17 ` Jason Gunthorpe via iommu 2021-11-15 17:54 ` Robin Murphy 2021-11-15 17:54 ` Robin Murphy 2021-11-15 18:19 ` Christoph Hellwig 2021-11-15 18:19 ` Christoph Hellwig 2021-11-15 18:44 ` Robin Murphy 2021-11-15 18:44 ` Robin Murphy 2021-11-15 19:22 ` Jason Gunthorpe via iommu 2021-11-15 19:22 ` Jason Gunthorpe 2021-11-15 20:58 ` Robin Murphy 2021-11-15 20:58 ` Robin Murphy 2021-11-15 21:19 ` Jason Gunthorpe via iommu 2021-11-15 21:19 ` Jason Gunthorpe 2021-11-15 20:48 ` Bjorn Helgaas 2021-11-15 20:48 ` Bjorn Helgaas 2021-11-15 22:17 ` Bjorn Helgaas 2021-11-15 22:17 ` Bjorn Helgaas 2021-11-16 6:05 ` Lu Baolu 2021-11-16 6:05 ` Lu Baolu 2021-11-15 2:05 ` [PATCH 04/11] PCI: portdrv: " Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 20:44 ` Bjorn Helgaas 2021-11-15 20:44 ` Bjorn Helgaas 2021-11-16 7:24 ` Lu Baolu 2021-11-16 7:24 ` Lu Baolu 2021-11-16 20:22 ` Bjorn Helgaas 2021-11-16 20:22 ` Bjorn Helgaas 2021-11-16 20:48 ` Jason Gunthorpe 2021-11-16 20:48 ` Jason Gunthorpe via iommu 2021-11-15 2:05 ` [PATCH 05/11] iommu: Add security context management for assigned devices Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 13:22 ` Christoph Hellwig 2021-11-15 13:22 ` Christoph Hellwig 2021-11-16 7:25 ` Lu Baolu 2021-11-16 7:25 ` Lu Baolu 2021-11-15 2:05 ` [PATCH 06/11] iommu: Expose group variants of dma ownership interfaces Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 13:27 ` Christoph Hellwig 2021-11-15 13:27 ` Christoph Hellwig 2021-11-16 9:42 ` Lu Baolu 2021-11-16 9:42 ` Lu Baolu 2021-11-15 2:05 ` [PATCH 07/11] vfio: Use DMA_OWNER_USER to declaim passthrough devices Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 2:05 ` [PATCH 08/11] vfio: Remove use of vfio_group_viable() Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 2:05 ` [PATCH 09/11] vfio: Delete the unbound_list Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 2:05 ` [PATCH 10/11] vfio: Remove iommu group notifier Lu Baolu 2021-11-15 2:05 ` Lu Baolu 2021-11-15 2:05 ` [PATCH 11/11] iommu: Remove iommu group changes notifier Lu Baolu 2021-11-15 2:05 ` Lu Baolu
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20211115020552.2378167-1-baolu.lu@linux.intel.com \ --to=baolu.lu@linux.intel.com \ --cc=alex.williamson@redhat.com \ --cc=ashok.raj@intel.com \ --cc=bhelgaas@google.com \ --cc=cohuck@redhat.com \ --cc=diana.craciun@oss.nxp.com \ --cc=eric.auger@redhat.com \ --cc=gregkh@linuxfoundation.org \ --cc=iommu@lists.linux-foundation.org \ --cc=jacob.jun.pan@intel.com \ --cc=jgg@nvidia.com \ --cc=joro@8bytes.org \ --cc=kch@nvidia.com \ --cc=kevin.tian@intel.com \ --cc=kvm@vger.kernel.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-pci@vger.kernel.org \ --cc=rafael@kernel.org \ --cc=will@kernel.org \ --cc=yi.l.liu@intel.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.