kernel.git/drivers/vfio/pci, branch linux-6.9.y

vfio/pci: Insert full vma on mmap'd MMIO fault

2024-07-25T07:53:27Z

[ Upstream commit d71a989cf5d961989c273093cdff2550acdde314 ] In order to improve performance of typical scenarios we can try to insert the entire vma on fault. This accelerates typical cases, such as when the MMIO region is DMA mapped by QEMU. The vfio_iommu_type1 driver will fault in the entire DMA mapped range through fixup_user_fault(). In synthetic testing, this improves the time required to walk a PCI BAR mapping from userspace by roughly 1/3rd. This is likely an interim solution until vmf_insert_pfn_{pmd,pud}() gain support for pfnmaps. Suggested-by: Yan Zhao Link: https://lore.kernel.org/all/Zl6XdUkt%2FzMMGOLF@yzhao56-desk.sh.intel.com/ Reviewed-by: Yan Zhao Link: https://lore.kernel.org/r/20240607035213.2054226-1-alex.williamson@redhat.com Signed-off-by: Alex Williamson Signed-off-by: Sasha Levin

vfio/pci: Use unmap_mapping_range()

2024-07-25T07:53:26Z

[ Upstream commit aac6db75a9fc2c7a6f73e152df8f15101dda38e6 ] With the vfio device fd tied to the address space of the pseudo fs inode, we can use the mm to track all vmas that might be mmap'ing device BARs, which removes our vma_list and all the complicated lock ordering necessary to manually zap each related vma. Note that we can no longer store the pfn in vm_pgoff if we want to use unmap_mapping_range() to zap a selective portion of the device fd corresponding to BAR mappings. This also converts our mmap fault handler to use vmf_insert_pfn() because we no longer have a vma_list to avoid the concurrency problem with io_remap_pfn_range(). The goal is to eventually use the vm_ops huge_fault handler to avoid the additional faulting overhead, but vmf_insert_pfn_{pmd,pud}() need to learn about pfnmaps first. Also, Jason notes that a race exists between unmap_mapping_range() and the fops mmap callback if we were to call io_remap_pfn_range() to populate the vma on mmap. Specifically, mmap_region() does call_mmap() before it does vma_link_file() which gives a window where the vma is populated but invisible to unmap_mapping_range(). Suggested-by: Jason Gunthorpe Reviewed-by: Jason Gunthorpe Reviewed-by: Kevin Tian Link: https://lore.kernel.org/r/20240530045236.1005864-3-alex.williamson@redhat.com Signed-off-by: Alex Williamson Signed-off-by: Sasha Levin

vfio/pci: Init the count variable in collecting hot-reset devices

2024-07-18T11:22:36Z

[ Upstream commit 5a88a3f67e37e39f933b38ebb4985ba5822e9eca ] The count variable is used without initialization, it results in mistakes in the device counting and crashes the userspace if the get hot reset info path is triggered. Fixes: f6944d4a0b87 ("vfio/pci: Collect hot-reset devices to local buffer") Link: https://bugzilla.kernel.org/show_bug.cgi?id=219010 Reported-by: Žilvinas Žaltiena Cc: Beld Zhang Signed-off-by: Yi Liu Reviewed-by: Kevin Tian Reviewed-by: Jason Gunthorpe Link: https://lore.kernel.org/r/20240710004150.319105-1-yi.l.liu@intel.com Signed-off-by: Alex Williamson Signed-off-by: Sasha Levin

vfio/pci: Collect hot-reset devices to local buffer

2024-06-27T11:52:19Z

[ Upstream commit f6944d4a0b87c16bc34ae589169e1ded3d4db08e ] Lockdep reports the below circular locking dependency issue. The mmap_lock acquisition while holding pci_bus_sem is due to the use of copy_to_user() from within a pci_walk_bus() callback. Building the devices array directly into the user buffer is only for convenience. Instead we can allocate a local buffer for the array, bounded by the number of devices on the bus/slot, fill the device information into this local buffer, then copy it into the user buffer outside the bus walk callback. ====================================================== WARNING: possible circular locking dependency detected 6.9.0-rc5+ #39 Not tainted ------------------------------------------------------ CPU 0/KVM/4113 is trying to acquire lock: ffff99a609ee18a8 (&vdev->vma_lock){+.+.}-{4:4}, at: vfio_pci_mmap_fault+0x35/0x1a0 [vfio_pci_core] but task is already holding lock: ffff99a243a052a0 (&mm->mmap_lock){++++}-{4:4}, at: vaddr_get_pfns+0x3f/0x170 [vfio_iommu_type1] which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #3 (&mm->mmap_lock){++++}-{4:4}: __lock_acquire+0x4e4/0xb90 lock_acquire+0xbc/0x2d0 __might_fault+0x5c/0x80 _copy_to_user+0x1e/0x60 vfio_pci_fill_devs+0x9f/0x130 [vfio_pci_core] vfio_pci_walk_wrapper+0x45/0x60 [vfio_pci_core] __pci_walk_bus+0x6b/0xb0 vfio_pci_ioctl_get_pci_hot_reset_info+0x10b/0x1d0 [vfio_pci_core] vfio_pci_core_ioctl+0x1cb/0x400 [vfio_pci_core] vfio_device_fops_unl_ioctl+0x7e/0x140 [vfio] __x64_sys_ioctl+0x8a/0xc0 do_syscall_64+0x8d/0x170 entry_SYSCALL_64_after_hwframe+0x76/0x7e -> #2 (pci_bus_sem){++++}-{4:4}: __lock_acquire+0x4e4/0xb90 lock_acquire+0xbc/0x2d0 down_read+0x3e/0x160 pci_bridge_wait_for_secondary_bus.part.0+0x33/0x2d0 pci_reset_bus+0xdd/0x160 vfio_pci_dev_set_hot_reset+0x256/0x270 [vfio_pci_core] vfio_pci_ioctl_pci_hot_reset_groups+0x1a3/0x280 [vfio_pci_core] vfio_pci_core_ioctl+0x3b5/0x400 [vfio_pci_core] vfio_device_fops_unl_ioctl+0x7e/0x140 [vfio] __x64_sys_ioctl+0x8a/0xc0 do_syscall_64+0x8d/0x170 entry_SYSCALL_64_after_hwframe+0x76/0x7e -> #1 (&vdev->memory_lock){+.+.}-{4:4}: __lock_acquire+0x4e4/0xb90 lock_acquire+0xbc/0x2d0 down_write+0x3b/0xc0 vfio_pci_zap_and_down_write_memory_lock+0x1c/0x30 [vfio_pci_core] vfio_basic_config_write+0x281/0x340 [vfio_pci_core] vfio_config_do_rw+0x1fa/0x300 [vfio_pci_core] vfio_pci_config_rw+0x75/0xe50 [vfio_pci_core] vfio_pci_rw+0xea/0x1a0 [vfio_pci_core] vfs_write+0xea/0x520 __x64_sys_pwrite64+0x90/0xc0 do_syscall_64+0x8d/0x170 entry_SYSCALL_64_after_hwframe+0x76/0x7e -> #0 (&vdev->vma_lock){+.+.}-{4:4}: check_prev_add+0xeb/0xcc0 validate_chain+0x465/0x530 __lock_acquire+0x4e4/0xb90 lock_acquire+0xbc/0x2d0 __mutex_lock+0x97/0xde0 vfio_pci_mmap_fault+0x35/0x1a0 [vfio_pci_core] __do_fault+0x31/0x160 do_pte_missing+0x65/0x3b0 __handle_mm_fault+0x303/0x720 handle_mm_fault+0x10f/0x460 fixup_user_fault+0x7f/0x1f0 follow_fault_pfn+0x66/0x1c0 [vfio_iommu_type1] vaddr_get_pfns+0xf2/0x170 [vfio_iommu_type1] vfio_pin_pages_remote+0x348/0x4e0 [vfio_iommu_type1] vfio_pin_map_dma+0xd2/0x330 [vfio_iommu_type1] vfio_dma_do_map+0x2c0/0x440 [vfio_iommu_type1] vfio_iommu_type1_ioctl+0xc5/0x1d0 [vfio_iommu_type1] __x64_sys_ioctl+0x8a/0xc0 do_syscall_64+0x8d/0x170 entry_SYSCALL_64_after_hwframe+0x76/0x7e other info that might help us debug this: Chain exists of: &vdev->vma_lock --> pci_bus_sem --> &mm->mmap_lock Possible unsafe locking scenario: block dm-0: the capability attribute has been deprecated. CPU0 CPU1 ---- ---- rlock(&mm->mmap_lock); lock(pci_bus_sem); lock(&mm->mmap_lock); lock(&vdev->vma_lock); *** DEADLOCK *** 2 locks held by CPU 0/KVM/4113: #0: ffff99a25f294888 (&iommu->lock#2){+.+.}-{4:4}, at: vfio_dma_do_map+0x60/0x440 [vfio_iommu_type1] #1: ffff99a243a052a0 (&mm->mmap_lock){++++}-{4:4}, at: vaddr_get_pfns+0x3f/0x170 [vfio_iommu_type1] stack backtrace: CPU: 1 PID: 4113 Comm: CPU 0/KVM Not tainted 6.9.0-rc5+ #39 Hardware name: Dell Inc. PowerEdge T640/04WYPY, BIOS 2.15.1 06/16/2022 Call Trace: dump_stack_lvl+0x64/0xa0 check_noncircular+0x131/0x150 check_prev_add+0xeb/0xcc0 ? add_chain_cache+0x10a/0x2f0 ? __lock_acquire+0x4e4/0xb90 validate_chain+0x465/0x530 __lock_acquire+0x4e4/0xb90 lock_acquire+0xbc/0x2d0 ? vfio_pci_mmap_fault+0x35/0x1a0 [vfio_pci_core] ? lock_is_held_type+0x9a/0x110 __mutex_lock+0x97/0xde0 ? vfio_pci_mmap_fault+0x35/0x1a0 [vfio_pci_core] ? lock_acquire+0xbc/0x2d0 ? vfio_pci_mmap_fault+0x35/0x1a0 [vfio_pci_core] ? find_held_lock+0x2b/0x80 ? vfio_pci_mmap_fault+0x35/0x1a0 [vfio_pci_core] vfio_pci_mmap_fault+0x35/0x1a0 [vfio_pci_core] __do_fault+0x31/0x160 do_pte_missing+0x65/0x3b0 __handle_mm_fault+0x303/0x720 handle_mm_fault+0x10f/0x460 fixup_user_fault+0x7f/0x1f0 follow_fault_pfn+0x66/0x1c0 [vfio_iommu_type1] vaddr_get_pfns+0xf2/0x170 [vfio_iommu_type1] vfio_pin_pages_remote+0x348/0x4e0 [vfio_iommu_type1] vfio_pin_map_dma+0xd2/0x330 [vfio_iommu_type1] vfio_dma_do_map+0x2c0/0x440 [vfio_iommu_type1] vfio_iommu_type1_ioctl+0xc5/0x1d0 [vfio_iommu_type1] __x64_sys_ioctl+0x8a/0xc0 do_syscall_64+0x8d/0x170 ? rcu_core+0x8d/0x250 ? __lock_release+0x5e/0x160 ? rcu_core+0x8d/0x250 ? lock_release+0x5f/0x120 ? sched_clock+0xc/0x30 ? sched_clock_cpu+0xb/0x190 ? irqtime_account_irq+0x40/0xc0 ? __local_bh_enable+0x54/0x60 ? __do_softirq+0x315/0x3ca ? lockdep_hardirqs_on_prepare.part.0+0x97/0x140 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f8300d0357b Code: ff ff ff 85 c0 79 9b 49 c7 c4 ff ff ff ff 5b 5d 4c 89 e0 41 5c c3 66 0f 1f 84 00 00 00 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 75 68 0f 00 f7 d8 64 89 01 48 RSP: 002b:00007f82ef3fb948 EFLAGS: 00000206 ORIG_RAX: 0000000000000010 RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f8300d0357b RDX: 00007f82ef3fb990 RSI: 0000000000003b71 RDI: 0000000000000023 RBP: 00007f82ef3fb9c0 R08: 0000000000000000 R09: 0000561b7e0bcac2 R10: 0000000000000000 R11: 0000000000000206 R12: 0000000000000000 R13: 0000000200000000 R14: 0000381800000000 R15: 0000000000000000 Reviewed-by: Jason Gunthorpe Link: https://lore.kernel.org/r/20240503143138.3562116-1-alex.williamson@redhat.com Signed-off-by: Alex Williamson Signed-off-by: Sasha Levin

vfio/pci: fix potential memory leak in vfio_intx_enable()

2024-06-12T09:39:13Z

[ Upstream commit 82b951e6fbd31d85ae7f4feb5f00ddd4c5d256e2 ] If vfio_irq_ctx_alloc() failed will lead to 'name' memory leak. Fixes: 18c198c96a81 ("vfio/pci: Create persistent INTx handler") Signed-off-by: Ye Bin Reviewed-by: Kevin Tian Acked-by: Reinette Chatre Link: https://lore.kernel.org/r/20240415015029.3699844-1-yebin10@huawei.com Signed-off-by: Alex Williamson Signed-off-by: Sasha Levin

VFIO: Add the SPR_DSA and SPR_IAX devices to the denylist

2024-05-17T10:18:09Z

commit 95feb3160eef0caa6018e175a5560b816aee8e79 upstream. Due to an erratum with the SPR_DSA and SPR_IAX devices, it is not secure to assign these devices to virtual machines. Add the PCI IDs of these devices to the VFIO denylist to ensure that this is handled appropriately by the VFIO subsystem. The SPR_DSA and SPR_IAX devices are on-SOC devices for the Sapphire Rapids (and related) family of products that perform data movement and compression. Signed-off-by: Arjan van de Ven Signed-off-by: Greg Kroah-Hartman

vfio/pci: Create persistent INTx handler

2024-03-11T19:08:52Z

A vulnerability exists where the eventfd for INTx signaling can be deconfigured, which unregisters the IRQ handler but still allows eventfds to be signaled with a NULL context through the SET_IRQS ioctl or through unmask irqfd if the device interrupt is pending. Ideally this could be solved with some additional locking; the igate mutex serializes the ioctl and config space accesses, and the interrupt handler is unregistered relative to the trigger, but the irqfd path runs asynchronous to those. The igate mutex cannot be acquired from the atomic context of the eventfd wake function. Disabling the irqfd relative to the eventfd registration is potentially incompatible with existing userspace. As a result, the solution implemented here moves configuration of the INTx interrupt handler to track the lifetime of the INTx context object and irq_type configuration, rather than registration of a particular trigger eventfd. Synchronization is added between the ioctl path and eventfd_signal() wrapper such that the eventfd trigger can be dynamically updated relative to in-flight interrupts or irqfd callbacks. Cc: Fixes: 89e1f7d4c66d ("vfio: Add PCI device driver") Reported-by: Reinette Chatre Reviewed-by: Kevin Tian Reviewed-by: Reinette Chatre Reviewed-by: Eric Auger Link: https://lore.kernel.org/r/20240308230557.805580-5-alex.williamson@redhat.com Signed-off-by: Alex Williamson

vfio/pci: Lock external INTx masking ops

2024-03-11T19:08:52Z

Mask operations through config space changes to DisINTx may race INTx configuration changes via ioctl. Create wrappers that add locking for paths outside of the core interrupt code. In particular, irq_type is updated holding igate, therefore testing is_intx() requires holding igate. For example clearing DisINTx from config space can otherwise race changes of the interrupt configuration. This aligns interfaces which may trigger the INTx eventfd into two camps, one side serialized by igate and the other only enabled while INTx is configured. A subsequent patch introduces synchronization for the latter flows. Cc: Fixes: 89e1f7d4c66d ("vfio: Add PCI device driver") Reported-by: Reinette Chatre Reviewed-by: Kevin Tian Reviewed-by: Reinette Chatre Reviewed-by: Eric Auger Link: https://lore.kernel.org/r/20240308230557.805580-3-alex.williamson@redhat.com Signed-off-by: Alex Williamson

vfio/pci: Disable auto-enable of exclusive INTx IRQ

2024-03-11T19:08:30Z

Currently for devices requiring masking at the irqchip for INTx, ie. devices without DisINTx support, the IRQ is enabled in request_irq() and subsequently disabled as necessary to align with the masked status flag. This presents a window where the interrupt could fire between these events, resulting in the IRQ incrementing the disable depth twice. This would be unrecoverable for a user since the masked flag prevents nested enables through vfio. Instead, invert the logic using IRQF_NO_AUTOEN such that exclusive INTx is never auto-enabled, then unmask as required. Cc: Fixes: 89e1f7d4c66d ("vfio: Add PCI device driver") Reviewed-by: Kevin Tian Reviewed-by: Eric Auger Link: https://lore.kernel.org/r/20240308230557.805580-2-alex.williamson@redhat.com Signed-off-by: Alex Williamson

vfio/pds: Refactor/simplify reset logic

2024-03-11T18:10:37Z

The current logic for handling resets is more complicated than it needs to be. The deferred_reset flag is used to indicate a reset is needed and the deferred_reset_state is the requested, post-reset, state. Also, the deferred_reset logic was added to vfio migration drivers to prevent a circular locking dependency with respect to mm_lock and state mutex. This is mainly because of the copy_to/from_user() functions(which takes mm_lock) invoked under state mutex. Remove all of the deferred reset logic and just pass the requested next state to pds_vfio_reset() so it can be used for VMM and DSC initiated resets. This removes the need for pds_vfio_state_mutex_lock(), so remove that and replace its use with a simple mutex_unlock(). Also, remove the reset_mutex as it's no longer needed since the state_mutex can be the driver's primary protector. Suggested-by: Shameer Kolothum Reviewed-by: Shannon Nelson Signed-off-by: Brett Creeley Link: https://lore.kernel.org/r/20240308182149.22036-3-brett.creeley@amd.com Signed-off-by: Alex Williamson