linux/drivers/iommu/iommufd/iommufd_private.h

738 lines
23 KiB
C
Raw Permalink Normal View History

/* SPDX-License-Identifier: GPL-2.0-only */
/* Copyright (c) 2021-2022, NVIDIA CORPORATION & AFFILIATES
*/
#ifndef __IOMMUFD_PRIVATE_H
#define __IOMMUFD_PRIVATE_H
#include <linux/iommu.h>
#include <linux/iommufd.h>
#include <linux/iova_bitmap.h>
#include <linux/maple_tree.h>
#include <linux/rwsem.h>
#include <linux/uaccess.h>
#include <linux/xarray.h>
#include <uapi/linux/iommufd.h>
#include "../iommu-priv.h"
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
struct iommu_domain;
struct iommu_group;
struct iommu_option;
struct iommufd_device;
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
struct iommufd_sw_msi_map {
struct list_head sw_msi_item;
phys_addr_t sw_msi_start;
phys_addr_t msi_addr;
unsigned int pgoff;
unsigned int id;
};
/* Bitmap of struct iommufd_sw_msi_map::id */
struct iommufd_sw_msi_maps {
DECLARE_BITMAP(bitmap, 64);
};
#ifdef CONFIG_IRQ_MSI_IOMMU
int iommufd_sw_msi_install(struct iommufd_ctx *ictx,
struct iommufd_hwpt_paging *hwpt_paging,
struct iommufd_sw_msi_map *msi_map);
#endif
struct iommufd_ctx {
struct file *file;
struct xarray objects;
struct xarray groups;
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
wait_queue_head_t destroy_wait;
struct rw_semaphore ioas_creation_lock;
struct maple_tree mt_mmap;
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
struct mutex sw_msi_lock;
struct list_head sw_msi_list;
unsigned int sw_msi_id;
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
u8 account_mode;
/* Compatibility with VFIO no iommu */
u8 no_iommu_mode;
iommufd: vfio container FD ioctl compatibility iommufd can directly implement the /dev/vfio/vfio container IOCTLs by mapping them into io_pagetable operations. A userspace application can test against iommufd and confirm compatibility then simply make a small change to open /dev/iommu instead of /dev/vfio/vfio. For testing purposes /dev/vfio/vfio can be symlinked to /dev/iommu and then all applications will use the compatibility path with no code changes. A later series allows /dev/vfio/vfio to be directly provided by iommufd, which allows the rlimit mode to work the same as well. This series just provides the iommufd side of compatibility. Actually linking this to VFIO_SET_CONTAINER is a followup series, with a link in the cover letter. Internally the compatibility API uses a normal IOAS object that, like vfio, is automatically allocated when the first device is attached. Userspace can also query or set this IOAS object directly using the IOMMU_VFIO_IOAS ioctl. This allows mixing and matching new iommufd only features while still using the VFIO style map/unmap ioctls. While this is enough to operate qemu, it has a few differences: - Resource limits rely on memory cgroups to bound what userspace can do instead of the module parameter dma_entry_limit. - VFIO P2P is not implemented. The DMABUF patches for vfio are a start at a solution where iommufd would import a special DMABUF. This is to avoid further propogating the follow_pfn() security problem. - A full audit for pedantic compatibility details (eg errnos, etc) has not yet been done - powerpc SPAPR is left out, as it is not connected to the iommu_domain framework. It seems interest in SPAPR is minimal as it is currently non-working in v6.1-rc1. They will have to convert to the iommu subsystem framework to enjoy iommfd. The following are not going to be implemented and we expect to remove them from VFIO type1: - SW access 'dirty tracking'. As discussed in the cover letter this will be done in VFIO. - VFIO_TYPE1_NESTING_IOMMU https://lore.kernel.org/all/0-v1-0093c9b0e345+19-vfio_no_nesting_jgg@nvidia.com/ - VFIO_DMA_MAP_FLAG_VADDR https://lore.kernel.org/all/Yz777bJZjTyLrHEQ@nvidia.com/ Link: https://lore.kernel.org/r/15-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:38 -04:00
struct iommufd_ioas *vfio_ioas;
};
/* Entry for iommufd_ctx::mt_mmap */
struct iommufd_mmap {
struct iommufd_object *owner;
/* Page-shifted start position in mt_mmap to validate vma->vm_pgoff */
unsigned long vm_pgoff;
/* Physical range for io_remap_pfn_range() */
phys_addr_t mmio_addr;
size_t length;
};
iommufd: PFN handling for iopt_pages The top of the data structure provides an IO Address Space (IOAS) that is similar to a VFIO container. The IOAS allows map/unmap of memory into ranges of IOVA called iopt_areas. Multiple IOMMU domains (IO page tables) and in-kernel accesses (like VFIO mdevs) can be attached to the IOAS to access the PFNs that those IOVA areas cover. The IO Address Space (IOAS) datastructure is composed of: - struct io_pagetable holding the IOVA map - struct iopt_areas representing populated portions of IOVA - struct iopt_pages representing the storage of PFNs - struct iommu_domain representing each IO page table in the system IOMMU - struct iopt_pages_access representing in-kernel accesses of PFNs (ie VFIO mdevs) - struct xarray pinned_pfns holding a list of pages pinned by in-kernel accesses This patch introduces the lowest part of the datastructure - the movement of PFNs in a tiered storage scheme: 1) iopt_pages::pinned_pfns xarray 2) Multiple iommu_domains 3) The origin of the PFNs, i.e. the userspace pointer PFN have to be copied between all combinations of tiers, depending on the configuration. The interface is an iterator called a 'pfn_reader' which determines which tier each PFN is stored and loads it into a list of PFNs held in a struct pfn_batch. Each step of the iterator will fill up the pfn_batch, then the caller can use the pfn_batch to send the PFNs to the required destination. Repeating this loop will read all the PFNs in an IOVA range. The pfn_reader and pfn_batch also keep track of the pinned page accounting. While PFNs are always stored and accessed as full PAGE_SIZE units the iommu_domain tier can store with a sub-page offset/length to support IOMMUs with a smaller IOPTE size than PAGE_SIZE. Link: https://lore.kernel.org/r/8-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:31 -04:00
/*
* The IOVA to PFN map. The map automatically copies the PFNs into multiple
* domains and permits sharing of PFNs between io_pagetable instances. This
* supports both a design where IOAS's are 1:1 with a domain (eg because the
* domain is HW customized), or where the IOAS is 1:N with multiple generic
* domains. The io_pagetable holds an interval tree of iopt_areas which point
* to shared iopt_pages which hold the pfns mapped to the page table.
*
* The locking order is domains_rwsem -> iova_rwsem -> pages::mutex
*/
struct io_pagetable {
struct rw_semaphore domains_rwsem;
struct xarray domains;
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
struct xarray access_list;
iommufd: PFN handling for iopt_pages The top of the data structure provides an IO Address Space (IOAS) that is similar to a VFIO container. The IOAS allows map/unmap of memory into ranges of IOVA called iopt_areas. Multiple IOMMU domains (IO page tables) and in-kernel accesses (like VFIO mdevs) can be attached to the IOAS to access the PFNs that those IOVA areas cover. The IO Address Space (IOAS) datastructure is composed of: - struct io_pagetable holding the IOVA map - struct iopt_areas representing populated portions of IOVA - struct iopt_pages representing the storage of PFNs - struct iommu_domain representing each IO page table in the system IOMMU - struct iopt_pages_access representing in-kernel accesses of PFNs (ie VFIO mdevs) - struct xarray pinned_pfns holding a list of pages pinned by in-kernel accesses This patch introduces the lowest part of the datastructure - the movement of PFNs in a tiered storage scheme: 1) iopt_pages::pinned_pfns xarray 2) Multiple iommu_domains 3) The origin of the PFNs, i.e. the userspace pointer PFN have to be copied between all combinations of tiers, depending on the configuration. The interface is an iterator called a 'pfn_reader' which determines which tier each PFN is stored and loads it into a list of PFNs held in a struct pfn_batch. Each step of the iterator will fill up the pfn_batch, then the caller can use the pfn_batch to send the PFNs to the required destination. Repeating this loop will read all the PFNs in an IOVA range. The pfn_reader and pfn_batch also keep track of the pinned page accounting. While PFNs are always stored and accessed as full PAGE_SIZE units the iommu_domain tier can store with a sub-page offset/length to support IOMMUs with a smaller IOPTE size than PAGE_SIZE. Link: https://lore.kernel.org/r/8-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:31 -04:00
unsigned int next_domain_id;
struct rw_semaphore iova_rwsem;
struct rb_root_cached area_itree;
/* IOVA that cannot become reserved, struct iopt_allowed */
struct rb_root_cached allowed_itree;
/* IOVA that cannot be allocated, struct iopt_reserved */
struct rb_root_cached reserved_itree;
u8 disable_large_pages;
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
unsigned long iova_alignment;
};
void iopt_init_table(struct io_pagetable *iopt);
void iopt_destroy_table(struct io_pagetable *iopt);
int iopt_get_pages(struct io_pagetable *iopt, unsigned long iova,
unsigned long length, struct list_head *pages_list);
void iopt_free_pages_list(struct list_head *pages_list);
enum {
IOPT_ALLOC_IOVA = 1 << 0,
iommufd: PFN handling for iopt_pages The top of the data structure provides an IO Address Space (IOAS) that is similar to a VFIO container. The IOAS allows map/unmap of memory into ranges of IOVA called iopt_areas. Multiple IOMMU domains (IO page tables) and in-kernel accesses (like VFIO mdevs) can be attached to the IOAS to access the PFNs that those IOVA areas cover. The IO Address Space (IOAS) datastructure is composed of: - struct io_pagetable holding the IOVA map - struct iopt_areas representing populated portions of IOVA - struct iopt_pages representing the storage of PFNs - struct iommu_domain representing each IO page table in the system IOMMU - struct iopt_pages_access representing in-kernel accesses of PFNs (ie VFIO mdevs) - struct xarray pinned_pfns holding a list of pages pinned by in-kernel accesses This patch introduces the lowest part of the datastructure - the movement of PFNs in a tiered storage scheme: 1) iopt_pages::pinned_pfns xarray 2) Multiple iommu_domains 3) The origin of the PFNs, i.e. the userspace pointer PFN have to be copied between all combinations of tiers, depending on the configuration. The interface is an iterator called a 'pfn_reader' which determines which tier each PFN is stored and loads it into a list of PFNs held in a struct pfn_batch. Each step of the iterator will fill up the pfn_batch, then the caller can use the pfn_batch to send the PFNs to the required destination. Repeating this loop will read all the PFNs in an IOVA range. The pfn_reader and pfn_batch also keep track of the pinned page accounting. While PFNs are always stored and accessed as full PAGE_SIZE units the iommu_domain tier can store with a sub-page offset/length to support IOMMUs with a smaller IOPTE size than PAGE_SIZE. Link: https://lore.kernel.org/r/8-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:31 -04:00
};
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
int iopt_map_user_pages(struct iommufd_ctx *ictx, struct io_pagetable *iopt,
unsigned long *iova, void __user *uptr,
unsigned long length, int iommu_prot,
unsigned int flags);
int iopt_map_file_pages(struct iommufd_ctx *ictx, struct io_pagetable *iopt,
unsigned long *iova, struct file *file,
unsigned long start, unsigned long length,
int iommu_prot, unsigned int flags);
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
int iopt_map_pages(struct io_pagetable *iopt, struct list_head *pages_list,
unsigned long length, unsigned long *dst_iova,
int iommu_prot, unsigned int flags);
int iopt_unmap_iova(struct io_pagetable *iopt, unsigned long iova,
unsigned long length, unsigned long *unmapped);
int iopt_unmap_all(struct io_pagetable *iopt, unsigned long *unmapped);
int iopt_read_and_clear_dirty_data(struct io_pagetable *iopt,
struct iommu_domain *domain,
unsigned long flags,
struct iommu_hwpt_get_dirty_bitmap *bitmap);
int iopt_set_dirty_tracking(struct io_pagetable *iopt,
struct iommu_domain *domain, bool enable);
void iommufd_access_notify_unmap(struct io_pagetable *iopt, unsigned long iova,
unsigned long length);
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
int iopt_table_add_domain(struct io_pagetable *iopt,
struct iommu_domain *domain);
void iopt_table_remove_domain(struct io_pagetable *iopt,
struct iommu_domain *domain);
int iopt_table_enforce_dev_resv_regions(struct io_pagetable *iopt,
struct device *dev,
phys_addr_t *sw_msi_start);
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
int iopt_set_allow_iova(struct io_pagetable *iopt,
struct rb_root_cached *allowed_iova);
int iopt_reserve_iova(struct io_pagetable *iopt, unsigned long start,
unsigned long last, void *owner);
void iopt_remove_reserved_iova(struct io_pagetable *iopt, void *owner);
int iopt_cut_iova(struct io_pagetable *iopt, unsigned long *iovas,
size_t num_iovas);
void iopt_enable_large_pages(struct io_pagetable *iopt);
int iopt_disable_large_pages(struct io_pagetable *iopt);
iommufd: PFN handling for iopt_pages The top of the data structure provides an IO Address Space (IOAS) that is similar to a VFIO container. The IOAS allows map/unmap of memory into ranges of IOVA called iopt_areas. Multiple IOMMU domains (IO page tables) and in-kernel accesses (like VFIO mdevs) can be attached to the IOAS to access the PFNs that those IOVA areas cover. The IO Address Space (IOAS) datastructure is composed of: - struct io_pagetable holding the IOVA map - struct iopt_areas representing populated portions of IOVA - struct iopt_pages representing the storage of PFNs - struct iommu_domain representing each IO page table in the system IOMMU - struct iopt_pages_access representing in-kernel accesses of PFNs (ie VFIO mdevs) - struct xarray pinned_pfns holding a list of pages pinned by in-kernel accesses This patch introduces the lowest part of the datastructure - the movement of PFNs in a tiered storage scheme: 1) iopt_pages::pinned_pfns xarray 2) Multiple iommu_domains 3) The origin of the PFNs, i.e. the userspace pointer PFN have to be copied between all combinations of tiers, depending on the configuration. The interface is an iterator called a 'pfn_reader' which determines which tier each PFN is stored and loads it into a list of PFNs held in a struct pfn_batch. Each step of the iterator will fill up the pfn_batch, then the caller can use the pfn_batch to send the PFNs to the required destination. Repeating this loop will read all the PFNs in an IOVA range. The pfn_reader and pfn_batch also keep track of the pinned page accounting. While PFNs are always stored and accessed as full PAGE_SIZE units the iommu_domain tier can store with a sub-page offset/length to support IOMMUs with a smaller IOPTE size than PAGE_SIZE. Link: https://lore.kernel.org/r/8-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:31 -04:00
struct iommufd_ucmd {
struct iommufd_ctx *ictx;
void __user *ubuffer;
u32 user_size;
void *cmd;
struct iommufd_object *new_obj;
};
iommufd: vfio container FD ioctl compatibility iommufd can directly implement the /dev/vfio/vfio container IOCTLs by mapping them into io_pagetable operations. A userspace application can test against iommufd and confirm compatibility then simply make a small change to open /dev/iommu instead of /dev/vfio/vfio. For testing purposes /dev/vfio/vfio can be symlinked to /dev/iommu and then all applications will use the compatibility path with no code changes. A later series allows /dev/vfio/vfio to be directly provided by iommufd, which allows the rlimit mode to work the same as well. This series just provides the iommufd side of compatibility. Actually linking this to VFIO_SET_CONTAINER is a followup series, with a link in the cover letter. Internally the compatibility API uses a normal IOAS object that, like vfio, is automatically allocated when the first device is attached. Userspace can also query or set this IOAS object directly using the IOMMU_VFIO_IOAS ioctl. This allows mixing and matching new iommufd only features while still using the VFIO style map/unmap ioctls. While this is enough to operate qemu, it has a few differences: - Resource limits rely on memory cgroups to bound what userspace can do instead of the module parameter dma_entry_limit. - VFIO P2P is not implemented. The DMABUF patches for vfio are a start at a solution where iommufd would import a special DMABUF. This is to avoid further propogating the follow_pfn() security problem. - A full audit for pedantic compatibility details (eg errnos, etc) has not yet been done - powerpc SPAPR is left out, as it is not connected to the iommu_domain framework. It seems interest in SPAPR is minimal as it is currently non-working in v6.1-rc1. They will have to convert to the iommu subsystem framework to enjoy iommfd. The following are not going to be implemented and we expect to remove them from VFIO type1: - SW access 'dirty tracking'. As discussed in the cover letter this will be done in VFIO. - VFIO_TYPE1_NESTING_IOMMU https://lore.kernel.org/all/0-v1-0093c9b0e345+19-vfio_no_nesting_jgg@nvidia.com/ - VFIO_DMA_MAP_FLAG_VADDR https://lore.kernel.org/all/Yz777bJZjTyLrHEQ@nvidia.com/ Link: https://lore.kernel.org/r/15-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:38 -04:00
int iommufd_vfio_ioctl(struct iommufd_ctx *ictx, unsigned int cmd,
unsigned long arg);
/* Copy the response in ucmd->cmd back to userspace. */
static inline int iommufd_ucmd_respond(struct iommufd_ucmd *ucmd,
size_t cmd_len)
{
if (copy_to_user(ucmd->ubuffer, ucmd->cmd,
min_t(size_t, ucmd->user_size, cmd_len)))
return -EFAULT;
return 0;
}
static inline bool iommufd_lock_obj(struct iommufd_object *obj)
{
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
if (!refcount_inc_not_zero(&obj->users))
return false;
if (!refcount_inc_not_zero(&obj->wait_cnt)) {
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
/*
* If the caller doesn't already have a ref on obj this must be
* called under the xa_lock. Otherwise the caller is holding a
* ref on users. Thus it cannot be one before this decrement.
*/
refcount_dec(&obj->users);
return false;
}
return true;
}
struct iommufd_object *iommufd_get_object(struct iommufd_ctx *ictx, u32 id,
enum iommufd_object_type type);
static inline void iommufd_put_object(struct iommufd_ctx *ictx,
struct iommufd_object *obj)
{
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
/*
* Users first, then wait_cnt so that REMOVE_WAIT never sees a spurious
* !0 users with a 0 wait_cnt.
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
*/
refcount_dec(&obj->users);
if (refcount_dec_and_test(&obj->wait_cnt))
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
wake_up_interruptible_all(&ictx->destroy_wait);
}
void iommufd_object_abort(struct iommufd_ctx *ictx, struct iommufd_object *obj);
void iommufd_object_abort_and_destroy(struct iommufd_ctx *ictx,
struct iommufd_object *obj);
void iommufd_object_finalize(struct iommufd_ctx *ictx,
struct iommufd_object *obj);
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
enum {
REMOVE_WAIT = BIT(0),
REMOVE_OBJ_TOMBSTONE = BIT(1),
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
};
int iommufd_object_remove(struct iommufd_ctx *ictx,
struct iommufd_object *to_destroy, u32 id,
unsigned int flags);
/*
* The caller holds a users refcount and wants to destroy the object. At this
* point the caller has no wait_cnt reference and at least the xarray will be
* holding one.
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
*/
iommufd: IOMMUFD_DESTROY should not increase the refcount syzkaller found a race where IOMMUFD_DESTROY increments the refcount: obj = iommufd_get_object(ucmd->ictx, cmd->id, IOMMUFD_OBJ_ANY); if (IS_ERR(obj)) return PTR_ERR(obj); iommufd_ref_to_users(obj); /* See iommufd_ref_to_users() */ if (!iommufd_object_destroy_user(ucmd->ictx, obj)) As part of the sequence to join the two existing primitives together. Allowing the refcount the be elevated without holding the destroy_rwsem violates the assumption that all temporary refcount elevations are protected by destroy_rwsem. Racing IOMMUFD_DESTROY with iommufd_object_destroy_user() will cause spurious failures: WARNING: CPU: 0 PID: 3076 at drivers/iommu/iommufd/device.c:477 iommufd_access_destroy+0x18/0x20 drivers/iommu/iommufd/device.c:478 Modules linked in: CPU: 0 PID: 3076 Comm: syz-executor.0 Not tainted 6.3.0-rc1-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/03/2023 RIP: 0010:iommufd_access_destroy+0x18/0x20 drivers/iommu/iommufd/device.c:477 Code: e8 3d 4e 00 00 84 c0 74 01 c3 0f 0b c3 0f 1f 44 00 00 f3 0f 1e fa 48 89 fe 48 8b bf a8 00 00 00 e8 1d 4e 00 00 84 c0 74 01 c3 <0f> 0b c3 0f 1f 44 00 00 41 57 41 56 41 55 4c 8d ae d0 00 00 00 41 RSP: 0018:ffffc90003067e08 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff888109ea0300 RCX: 0000000000000000 RDX: 0000000000000001 RSI: 0000000000000000 RDI: 00000000ffffffff RBP: 0000000000000004 R08: 0000000000000000 R09: ffff88810bbb3500 R10: ffff88810bbb3e48 R11: 0000000000000000 R12: ffffc90003067e88 R13: ffffc90003067ea8 R14: ffff888101249800 R15: 00000000fffffffe FS: 00007ff7254fe6c0(0000) GS:ffff888237c00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000555557262da8 CR3: 000000010a6fd000 CR4: 0000000000350ef0 Call Trace: <TASK> iommufd_test_create_access drivers/iommu/iommufd/selftest.c:596 [inline] iommufd_test+0x71c/0xcf0 drivers/iommu/iommufd/selftest.c:813 iommufd_fops_ioctl+0x10f/0x1b0 drivers/iommu/iommufd/main.c:337 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:870 [inline] __se_sys_ioctl fs/ioctl.c:856 [inline] __x64_sys_ioctl+0x84/0xc0 fs/ioctl.c:856 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0x80 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd The solution is to not increment the refcount on the IOMMUFD_DESTROY path at all. Instead use the xa_lock to serialize everything. The refcount check == 1 and xa_erase can be done under a single critical region. This avoids the need for any refcount incrementing. It has the downside that if userspace races destroy with other operations it will get an EBUSY instead of waiting, but this is kind of racing is already dangerous. Fixes: 2ff4bed7fee7 ("iommufd: File descriptor, context, kconfig and makefiles") Link: https://lore.kernel.org/r/2-v1-85aacb2af554+bc-iommufd_syz3_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reported-by: syzbot+7574ebfe589049630608@syzkaller.appspotmail.com Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-07-25 16:05:49 -03:00
static inline void iommufd_object_destroy_user(struct iommufd_ctx *ictx,
struct iommufd_object *obj)
{
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
int ret;
ret = iommufd_object_remove(ictx, obj, obj->id, REMOVE_WAIT);
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
/*
* If there is a bug and we couldn't destroy the object then we did put
* back the caller's users refcount and will eventually try to free it
* again during close.
*/
WARN_ON(ret);
}
/*
* Similar to iommufd_object_destroy_user(), except that the object ID is left
* reserved/tombstoned.
*/
static inline void iommufd_object_tombstone_user(struct iommufd_ctx *ictx,
struct iommufd_object *obj)
{
int ret;
ret = iommufd_object_remove(ictx, obj, obj->id,
REMOVE_WAIT | REMOVE_OBJ_TOMBSTONE);
/*
* If there is a bug and we couldn't destroy the object then we did put
* back the caller's users refcount and will eventually try to free it
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
* again during close.
*/
WARN_ON(ret);
iommufd: IOMMUFD_DESTROY should not increase the refcount syzkaller found a race where IOMMUFD_DESTROY increments the refcount: obj = iommufd_get_object(ucmd->ictx, cmd->id, IOMMUFD_OBJ_ANY); if (IS_ERR(obj)) return PTR_ERR(obj); iommufd_ref_to_users(obj); /* See iommufd_ref_to_users() */ if (!iommufd_object_destroy_user(ucmd->ictx, obj)) As part of the sequence to join the two existing primitives together. Allowing the refcount the be elevated without holding the destroy_rwsem violates the assumption that all temporary refcount elevations are protected by destroy_rwsem. Racing IOMMUFD_DESTROY with iommufd_object_destroy_user() will cause spurious failures: WARNING: CPU: 0 PID: 3076 at drivers/iommu/iommufd/device.c:477 iommufd_access_destroy+0x18/0x20 drivers/iommu/iommufd/device.c:478 Modules linked in: CPU: 0 PID: 3076 Comm: syz-executor.0 Not tainted 6.3.0-rc1-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/03/2023 RIP: 0010:iommufd_access_destroy+0x18/0x20 drivers/iommu/iommufd/device.c:477 Code: e8 3d 4e 00 00 84 c0 74 01 c3 0f 0b c3 0f 1f 44 00 00 f3 0f 1e fa 48 89 fe 48 8b bf a8 00 00 00 e8 1d 4e 00 00 84 c0 74 01 c3 <0f> 0b c3 0f 1f 44 00 00 41 57 41 56 41 55 4c 8d ae d0 00 00 00 41 RSP: 0018:ffffc90003067e08 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff888109ea0300 RCX: 0000000000000000 RDX: 0000000000000001 RSI: 0000000000000000 RDI: 00000000ffffffff RBP: 0000000000000004 R08: 0000000000000000 R09: ffff88810bbb3500 R10: ffff88810bbb3e48 R11: 0000000000000000 R12: ffffc90003067e88 R13: ffffc90003067ea8 R14: ffff888101249800 R15: 00000000fffffffe FS: 00007ff7254fe6c0(0000) GS:ffff888237c00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000555557262da8 CR3: 000000010a6fd000 CR4: 0000000000350ef0 Call Trace: <TASK> iommufd_test_create_access drivers/iommu/iommufd/selftest.c:596 [inline] iommufd_test+0x71c/0xcf0 drivers/iommu/iommufd/selftest.c:813 iommufd_fops_ioctl+0x10f/0x1b0 drivers/iommu/iommufd/main.c:337 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:870 [inline] __se_sys_ioctl fs/ioctl.c:856 [inline] __x64_sys_ioctl+0x84/0xc0 fs/ioctl.c:856 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0x80 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd The solution is to not increment the refcount on the IOMMUFD_DESTROY path at all. Instead use the xa_lock to serialize everything. The refcount check == 1 and xa_erase can be done under a single critical region. This avoids the need for any refcount incrementing. It has the downside that if userspace races destroy with other operations it will get an EBUSY instead of waiting, but this is kind of racing is already dangerous. Fixes: 2ff4bed7fee7 ("iommufd: File descriptor, context, kconfig and makefiles") Link: https://lore.kernel.org/r/2-v1-85aacb2af554+bc-iommufd_syz3_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reported-by: syzbot+7574ebfe589049630608@syzkaller.appspotmail.com Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-07-25 16:05:49 -03:00
}
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
/*
* The HWPT allocated by autodomains is used in possibly many devices and
* is automatically destroyed when its refcount reaches zero.
*
* If userspace uses the HWPT manually, even for a short term, then it will
* disrupt this refcounting and the auto-free in the kernel will not work.
* Userspace that tries to use the automatically allocated HWPT must be careful
* to ensure that it is consistently destroyed, eg by not racing accesses
* and by not attaching an automatic HWPT to a device manually.
*/
static inline void
iommufd_object_put_and_try_destroy(struct iommufd_ctx *ictx,
struct iommufd_object *obj)
iommufd: IOMMUFD_DESTROY should not increase the refcount syzkaller found a race where IOMMUFD_DESTROY increments the refcount: obj = iommufd_get_object(ucmd->ictx, cmd->id, IOMMUFD_OBJ_ANY); if (IS_ERR(obj)) return PTR_ERR(obj); iommufd_ref_to_users(obj); /* See iommufd_ref_to_users() */ if (!iommufd_object_destroy_user(ucmd->ictx, obj)) As part of the sequence to join the two existing primitives together. Allowing the refcount the be elevated without holding the destroy_rwsem violates the assumption that all temporary refcount elevations are protected by destroy_rwsem. Racing IOMMUFD_DESTROY with iommufd_object_destroy_user() will cause spurious failures: WARNING: CPU: 0 PID: 3076 at drivers/iommu/iommufd/device.c:477 iommufd_access_destroy+0x18/0x20 drivers/iommu/iommufd/device.c:478 Modules linked in: CPU: 0 PID: 3076 Comm: syz-executor.0 Not tainted 6.3.0-rc1-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/03/2023 RIP: 0010:iommufd_access_destroy+0x18/0x20 drivers/iommu/iommufd/device.c:477 Code: e8 3d 4e 00 00 84 c0 74 01 c3 0f 0b c3 0f 1f 44 00 00 f3 0f 1e fa 48 89 fe 48 8b bf a8 00 00 00 e8 1d 4e 00 00 84 c0 74 01 c3 <0f> 0b c3 0f 1f 44 00 00 41 57 41 56 41 55 4c 8d ae d0 00 00 00 41 RSP: 0018:ffffc90003067e08 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff888109ea0300 RCX: 0000000000000000 RDX: 0000000000000001 RSI: 0000000000000000 RDI: 00000000ffffffff RBP: 0000000000000004 R08: 0000000000000000 R09: ffff88810bbb3500 R10: ffff88810bbb3e48 R11: 0000000000000000 R12: ffffc90003067e88 R13: ffffc90003067ea8 R14: ffff888101249800 R15: 00000000fffffffe FS: 00007ff7254fe6c0(0000) GS:ffff888237c00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000555557262da8 CR3: 000000010a6fd000 CR4: 0000000000350ef0 Call Trace: <TASK> iommufd_test_create_access drivers/iommu/iommufd/selftest.c:596 [inline] iommufd_test+0x71c/0xcf0 drivers/iommu/iommufd/selftest.c:813 iommufd_fops_ioctl+0x10f/0x1b0 drivers/iommu/iommufd/main.c:337 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:870 [inline] __se_sys_ioctl fs/ioctl.c:856 [inline] __x64_sys_ioctl+0x84/0xc0 fs/ioctl.c:856 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0x80 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd The solution is to not increment the refcount on the IOMMUFD_DESTROY path at all. Instead use the xa_lock to serialize everything. The refcount check == 1 and xa_erase can be done under a single critical region. This avoids the need for any refcount incrementing. It has the downside that if userspace races destroy with other operations it will get an EBUSY instead of waiting, but this is kind of racing is already dangerous. Fixes: 2ff4bed7fee7 ("iommufd: File descriptor, context, kconfig and makefiles") Link: https://lore.kernel.org/r/2-v1-85aacb2af554+bc-iommufd_syz3_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reported-by: syzbot+7574ebfe589049630608@syzkaller.appspotmail.com Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-07-25 16:05:49 -03:00
{
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
iommufd_object_remove(ictx, obj, obj->id, 0);
iommufd: IOMMUFD_DESTROY should not increase the refcount syzkaller found a race where IOMMUFD_DESTROY increments the refcount: obj = iommufd_get_object(ucmd->ictx, cmd->id, IOMMUFD_OBJ_ANY); if (IS_ERR(obj)) return PTR_ERR(obj); iommufd_ref_to_users(obj); /* See iommufd_ref_to_users() */ if (!iommufd_object_destroy_user(ucmd->ictx, obj)) As part of the sequence to join the two existing primitives together. Allowing the refcount the be elevated without holding the destroy_rwsem violates the assumption that all temporary refcount elevations are protected by destroy_rwsem. Racing IOMMUFD_DESTROY with iommufd_object_destroy_user() will cause spurious failures: WARNING: CPU: 0 PID: 3076 at drivers/iommu/iommufd/device.c:477 iommufd_access_destroy+0x18/0x20 drivers/iommu/iommufd/device.c:478 Modules linked in: CPU: 0 PID: 3076 Comm: syz-executor.0 Not tainted 6.3.0-rc1-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/03/2023 RIP: 0010:iommufd_access_destroy+0x18/0x20 drivers/iommu/iommufd/device.c:477 Code: e8 3d 4e 00 00 84 c0 74 01 c3 0f 0b c3 0f 1f 44 00 00 f3 0f 1e fa 48 89 fe 48 8b bf a8 00 00 00 e8 1d 4e 00 00 84 c0 74 01 c3 <0f> 0b c3 0f 1f 44 00 00 41 57 41 56 41 55 4c 8d ae d0 00 00 00 41 RSP: 0018:ffffc90003067e08 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff888109ea0300 RCX: 0000000000000000 RDX: 0000000000000001 RSI: 0000000000000000 RDI: 00000000ffffffff RBP: 0000000000000004 R08: 0000000000000000 R09: ffff88810bbb3500 R10: ffff88810bbb3e48 R11: 0000000000000000 R12: ffffc90003067e88 R13: ffffc90003067ea8 R14: ffff888101249800 R15: 00000000fffffffe FS: 00007ff7254fe6c0(0000) GS:ffff888237c00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000555557262da8 CR3: 000000010a6fd000 CR4: 0000000000350ef0 Call Trace: <TASK> iommufd_test_create_access drivers/iommu/iommufd/selftest.c:596 [inline] iommufd_test+0x71c/0xcf0 drivers/iommu/iommufd/selftest.c:813 iommufd_fops_ioctl+0x10f/0x1b0 drivers/iommu/iommufd/main.c:337 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:870 [inline] __se_sys_ioctl fs/ioctl.c:856 [inline] __x64_sys_ioctl+0x84/0xc0 fs/ioctl.c:856 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0x80 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd The solution is to not increment the refcount on the IOMMUFD_DESTROY path at all. Instead use the xa_lock to serialize everything. The refcount check == 1 and xa_erase can be done under a single critical region. This avoids the need for any refcount incrementing. It has the downside that if userspace races destroy with other operations it will get an EBUSY instead of waiting, but this is kind of racing is already dangerous. Fixes: 2ff4bed7fee7 ("iommufd: File descriptor, context, kconfig and makefiles") Link: https://lore.kernel.org/r/2-v1-85aacb2af554+bc-iommufd_syz3_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reported-by: syzbot+7574ebfe589049630608@syzkaller.appspotmail.com Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-07-25 16:05:49 -03:00
}
/*
* Callers of these normal object allocators must call iommufd_object_finalize()
* to finalize the object, or call iommufd_object_abort_and_destroy() to revert
* the allocation.
*/
struct iommufd_object *_iommufd_object_alloc(struct iommufd_ctx *ictx,
size_t size,
enum iommufd_object_type type);
#define __iommufd_object_alloc(ictx, ptr, type, obj) \
container_of(_iommufd_object_alloc( \
ictx, \
sizeof(*(ptr)) + BUILD_BUG_ON_ZERO( \
offsetof(typeof(*(ptr)), \
obj) != 0), \
type), \
typeof(*(ptr)), obj)
#define iommufd_object_alloc(ictx, ptr, type) \
__iommufd_object_alloc(ictx, ptr, type, obj)
/*
* Callers of these _ucmd allocators should not call iommufd_object_finalize()
* or iommufd_object_abort_and_destroy(), as the core automatically does that.
*/
struct iommufd_object *
_iommufd_object_alloc_ucmd(struct iommufd_ucmd *ucmd, size_t size,
enum iommufd_object_type type);
#define __iommufd_object_alloc_ucmd(ucmd, ptr, type, obj) \
container_of(_iommufd_object_alloc_ucmd( \
ucmd, \
sizeof(*(ptr)) + BUILD_BUG_ON_ZERO( \
offsetof(typeof(*(ptr)), \
obj) != 0), \
type), \
typeof(*(ptr)), obj)
#define iommufd_object_alloc_ucmd(ucmd, ptr, type) \
__iommufd_object_alloc_ucmd(ucmd, ptr, type, obj)
/*
* The IO Address Space (IOAS) pagetable is a virtual page table backed by the
* io_pagetable object. It is a user controlled mapping of IOVA -> PFNs. The
* mapping is copied into all of the associated domains and made available to
* in-kernel users.
iommufd: Add a HW pagetable object The hw_pagetable object exposes the internal struct iommu_domain's to userspace. An iommu_domain is required when any DMA device attaches to an IOAS to control the io page table through the iommu driver. For compatibility with VFIO the hw_pagetable is automatically created when a DMA device is attached to the IOAS. If a compatible iommu_domain already exists then the hw_pagetable associated with it is used for the attachment. In the initial series there is no iommufd uAPI for the hw_pagetable object. The next patch provides driver facing APIs for IO page table attachment that allows drivers to accept either an IOAS or a hw_pagetable ID and for the driver to return the hw_pagetable ID that was auto-selected from an IOAS. The expectation is the driver will provide uAPI through its own FD for attaching its device to iommufd. This allows userspace to learn the mapping of devices to iommu_domains and to override the automatic attachment. The future HW specific interface will allow userspace to create hw_pagetable objects using iommu_domains with IOMMU driver specific parameters. This infrastructure will allow linking those domains to IOAS's and devices. Link: https://lore.kernel.org/r/12-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:35 -04:00
*
* Every iommu_domain that is created is wrapped in a iommufd_hw_pagetable
* object. When we go to attach a device to an IOAS we need to get an
* iommu_domain and wrapping iommufd_hw_pagetable for it.
*
* An iommu_domain & iommfd_hw_pagetable will be automatically selected
* for a device based on the hwpt_list. If no suitable iommu_domain
* is found a new iommu_domain will be created.
*/
struct iommufd_ioas {
struct iommufd_object obj;
struct io_pagetable iopt;
iommufd: Add a HW pagetable object The hw_pagetable object exposes the internal struct iommu_domain's to userspace. An iommu_domain is required when any DMA device attaches to an IOAS to control the io page table through the iommu driver. For compatibility with VFIO the hw_pagetable is automatically created when a DMA device is attached to the IOAS. If a compatible iommu_domain already exists then the hw_pagetable associated with it is used for the attachment. In the initial series there is no iommufd uAPI for the hw_pagetable object. The next patch provides driver facing APIs for IO page table attachment that allows drivers to accept either an IOAS or a hw_pagetable ID and for the driver to return the hw_pagetable ID that was auto-selected from an IOAS. The expectation is the driver will provide uAPI through its own FD for attaching its device to iommufd. This allows userspace to learn the mapping of devices to iommu_domains and to override the automatic attachment. The future HW specific interface will allow userspace to create hw_pagetable objects using iommu_domains with IOMMU driver specific parameters. This infrastructure will allow linking those domains to IOAS's and devices. Link: https://lore.kernel.org/r/12-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:35 -04:00
struct mutex mutex;
struct list_head hwpt_list;
};
static inline struct iommufd_ioas *iommufd_get_ioas(struct iommufd_ctx *ictx,
u32 id)
{
return container_of(iommufd_get_object(ictx, id, IOMMUFD_OBJ_IOAS),
struct iommufd_ioas, obj);
}
struct iommufd_ioas *iommufd_ioas_alloc(struct iommufd_ctx *ictx);
int iommufd_ioas_alloc_ioctl(struct iommufd_ucmd *ucmd);
void iommufd_ioas_destroy(struct iommufd_object *obj);
int iommufd_ioas_iova_ranges(struct iommufd_ucmd *ucmd);
int iommufd_ioas_allow_iovas(struct iommufd_ucmd *ucmd);
int iommufd_ioas_map(struct iommufd_ucmd *ucmd);
int iommufd_ioas_map_file(struct iommufd_ucmd *ucmd);
int iommufd_ioas_change_process(struct iommufd_ucmd *ucmd);
int iommufd_ioas_copy(struct iommufd_ucmd *ucmd);
int iommufd_ioas_unmap(struct iommufd_ucmd *ucmd);
int iommufd_ioas_option(struct iommufd_ucmd *ucmd);
int iommufd_option_rlimit_mode(struct iommu_option *cmd,
struct iommufd_ctx *ictx);
iommufd: vfio container FD ioctl compatibility iommufd can directly implement the /dev/vfio/vfio container IOCTLs by mapping them into io_pagetable operations. A userspace application can test against iommufd and confirm compatibility then simply make a small change to open /dev/iommu instead of /dev/vfio/vfio. For testing purposes /dev/vfio/vfio can be symlinked to /dev/iommu and then all applications will use the compatibility path with no code changes. A later series allows /dev/vfio/vfio to be directly provided by iommufd, which allows the rlimit mode to work the same as well. This series just provides the iommufd side of compatibility. Actually linking this to VFIO_SET_CONTAINER is a followup series, with a link in the cover letter. Internally the compatibility API uses a normal IOAS object that, like vfio, is automatically allocated when the first device is attached. Userspace can also query or set this IOAS object directly using the IOMMU_VFIO_IOAS ioctl. This allows mixing and matching new iommufd only features while still using the VFIO style map/unmap ioctls. While this is enough to operate qemu, it has a few differences: - Resource limits rely on memory cgroups to bound what userspace can do instead of the module parameter dma_entry_limit. - VFIO P2P is not implemented. The DMABUF patches for vfio are a start at a solution where iommufd would import a special DMABUF. This is to avoid further propogating the follow_pfn() security problem. - A full audit for pedantic compatibility details (eg errnos, etc) has not yet been done - powerpc SPAPR is left out, as it is not connected to the iommu_domain framework. It seems interest in SPAPR is minimal as it is currently non-working in v6.1-rc1. They will have to convert to the iommu subsystem framework to enjoy iommfd. The following are not going to be implemented and we expect to remove them from VFIO type1: - SW access 'dirty tracking'. As discussed in the cover letter this will be done in VFIO. - VFIO_TYPE1_NESTING_IOMMU https://lore.kernel.org/all/0-v1-0093c9b0e345+19-vfio_no_nesting_jgg@nvidia.com/ - VFIO_DMA_MAP_FLAG_VADDR https://lore.kernel.org/all/Yz777bJZjTyLrHEQ@nvidia.com/ Link: https://lore.kernel.org/r/15-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:38 -04:00
int iommufd_vfio_ioas(struct iommufd_ucmd *ucmd);
int iommufd_check_iova_range(struct io_pagetable *iopt,
struct iommu_hwpt_get_dirty_bitmap *bitmap);
iommufd: vfio container FD ioctl compatibility iommufd can directly implement the /dev/vfio/vfio container IOCTLs by mapping them into io_pagetable operations. A userspace application can test against iommufd and confirm compatibility then simply make a small change to open /dev/iommu instead of /dev/vfio/vfio. For testing purposes /dev/vfio/vfio can be symlinked to /dev/iommu and then all applications will use the compatibility path with no code changes. A later series allows /dev/vfio/vfio to be directly provided by iommufd, which allows the rlimit mode to work the same as well. This series just provides the iommufd side of compatibility. Actually linking this to VFIO_SET_CONTAINER is a followup series, with a link in the cover letter. Internally the compatibility API uses a normal IOAS object that, like vfio, is automatically allocated when the first device is attached. Userspace can also query or set this IOAS object directly using the IOMMU_VFIO_IOAS ioctl. This allows mixing and matching new iommufd only features while still using the VFIO style map/unmap ioctls. While this is enough to operate qemu, it has a few differences: - Resource limits rely on memory cgroups to bound what userspace can do instead of the module parameter dma_entry_limit. - VFIO P2P is not implemented. The DMABUF patches for vfio are a start at a solution where iommufd would import a special DMABUF. This is to avoid further propogating the follow_pfn() security problem. - A full audit for pedantic compatibility details (eg errnos, etc) has not yet been done - powerpc SPAPR is left out, as it is not connected to the iommu_domain framework. It seems interest in SPAPR is minimal as it is currently non-working in v6.1-rc1. They will have to convert to the iommu subsystem framework to enjoy iommfd. The following are not going to be implemented and we expect to remove them from VFIO type1: - SW access 'dirty tracking'. As discussed in the cover letter this will be done in VFIO. - VFIO_TYPE1_NESTING_IOMMU https://lore.kernel.org/all/0-v1-0093c9b0e345+19-vfio_no_nesting_jgg@nvidia.com/ - VFIO_DMA_MAP_FLAG_VADDR https://lore.kernel.org/all/Yz777bJZjTyLrHEQ@nvidia.com/ Link: https://lore.kernel.org/r/15-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:38 -04:00
iommufd: Add a HW pagetable object The hw_pagetable object exposes the internal struct iommu_domain's to userspace. An iommu_domain is required when any DMA device attaches to an IOAS to control the io page table through the iommu driver. For compatibility with VFIO the hw_pagetable is automatically created when a DMA device is attached to the IOAS. If a compatible iommu_domain already exists then the hw_pagetable associated with it is used for the attachment. In the initial series there is no iommufd uAPI for the hw_pagetable object. The next patch provides driver facing APIs for IO page table attachment that allows drivers to accept either an IOAS or a hw_pagetable ID and for the driver to return the hw_pagetable ID that was auto-selected from an IOAS. The expectation is the driver will provide uAPI through its own FD for attaching its device to iommufd. This allows userspace to learn the mapping of devices to iommu_domains and to override the automatic attachment. The future HW specific interface will allow userspace to create hw_pagetable objects using iommu_domains with IOMMU driver specific parameters. This infrastructure will allow linking those domains to IOAS's and devices. Link: https://lore.kernel.org/r/12-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:35 -04:00
/*
* A HW pagetable is called an iommu_domain inside the kernel. This user object
* allows directly creating and inspecting the domains. Domains that have kernel
* owned page tables will be associated with an iommufd_ioas that provides the
* IOVA to PFN map.
*/
struct iommufd_hw_pagetable {
struct iommufd_object obj;
struct iommu_domain *domain;
struct iommufd_fault *fault;
bool pasid_compat : 1;
};
struct iommufd_hwpt_paging {
struct iommufd_hw_pagetable common;
struct iommufd_ioas *ioas;
iommufd: Add a HW pagetable object The hw_pagetable object exposes the internal struct iommu_domain's to userspace. An iommu_domain is required when any DMA device attaches to an IOAS to control the io page table through the iommu driver. For compatibility with VFIO the hw_pagetable is automatically created when a DMA device is attached to the IOAS. If a compatible iommu_domain already exists then the hw_pagetable associated with it is used for the attachment. In the initial series there is no iommufd uAPI for the hw_pagetable object. The next patch provides driver facing APIs for IO page table attachment that allows drivers to accept either an IOAS or a hw_pagetable ID and for the driver to return the hw_pagetable ID that was auto-selected from an IOAS. The expectation is the driver will provide uAPI through its own FD for attaching its device to iommufd. This allows userspace to learn the mapping of devices to iommu_domains and to override the automatic attachment. The future HW specific interface will allow userspace to create hw_pagetable objects using iommu_domains with IOMMU driver specific parameters. This infrastructure will allow linking those domains to IOAS's and devices. Link: https://lore.kernel.org/r/12-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:35 -04:00
bool auto_domain : 1;
bool enforce_cache_coherency : 1;
iommufd: Add a nested HW pagetable object IOMMU_HWPT_ALLOC already supports iommu_domain allocation for usersapce. But it can only allocate a hw_pagetable that associates to a given IOAS, i.e. only a kernel-managed hw_pagetable of IOMMUFD_OBJ_HWPT_PAGING type. IOMMU drivers can now support user-managed hw_pagetables, for two-stage translation use cases that require user data input from the user space. Add a new IOMMUFD_OBJ_HWPT_NESTED type with its abort/destroy(). Pair it with a new iommufd_hwpt_nested structure and its to_hwpt_nested() helper. Update the to_hwpt_paging() helper, so a NESTED-type hw_pagetable can be handled in the callers, for example iommufd_hw_pagetable_enforce_rr(). Screen the inputs including the parent PAGING-type hw_pagetable that has a need of a new nest_parent flag in the iommufd_hwpt_paging structure. Extend the IOMMU_HWPT_ALLOC ioctl to accept an IOMMU driver specific data input which is tagged by the enum iommu_hwpt_data_type. Also, update the @pt_id to accept hwpt_id too besides an ioas_id. Then, use them to allocate a hw_pagetable of IOMMUFD_OBJ_HWPT_NESTED type using the iommufd_hw_pagetable_alloc_nested() allocator. Link: https://lore.kernel.org/r/20231026043938.63898-8-yi.l.liu@intel.com Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Co-developed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-10-25 21:39:35 -07:00
bool nest_parent : 1;
iommufd: Add a HW pagetable object The hw_pagetable object exposes the internal struct iommu_domain's to userspace. An iommu_domain is required when any DMA device attaches to an IOAS to control the io page table through the iommu driver. For compatibility with VFIO the hw_pagetable is automatically created when a DMA device is attached to the IOAS. If a compatible iommu_domain already exists then the hw_pagetable associated with it is used for the attachment. In the initial series there is no iommufd uAPI for the hw_pagetable object. The next patch provides driver facing APIs for IO page table attachment that allows drivers to accept either an IOAS or a hw_pagetable ID and for the driver to return the hw_pagetable ID that was auto-selected from an IOAS. The expectation is the driver will provide uAPI through its own FD for attaching its device to iommufd. This allows userspace to learn the mapping of devices to iommu_domains and to override the automatic attachment. The future HW specific interface will allow userspace to create hw_pagetable objects using iommu_domains with IOMMU driver specific parameters. This infrastructure will allow linking those domains to IOAS's and devices. Link: https://lore.kernel.org/r/12-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:35 -04:00
/* Head at iommufd_ioas::hwpt_list */
struct list_head hwpt_item;
struct iommufd_sw_msi_maps present_sw_msi;
iommufd: Add a HW pagetable object The hw_pagetable object exposes the internal struct iommu_domain's to userspace. An iommu_domain is required when any DMA device attaches to an IOAS to control the io page table through the iommu driver. For compatibility with VFIO the hw_pagetable is automatically created when a DMA device is attached to the IOAS. If a compatible iommu_domain already exists then the hw_pagetable associated with it is used for the attachment. In the initial series there is no iommufd uAPI for the hw_pagetable object. The next patch provides driver facing APIs for IO page table attachment that allows drivers to accept either an IOAS or a hw_pagetable ID and for the driver to return the hw_pagetable ID that was auto-selected from an IOAS. The expectation is the driver will provide uAPI through its own FD for attaching its device to iommufd. This allows userspace to learn the mapping of devices to iommu_domains and to override the automatic attachment. The future HW specific interface will allow userspace to create hw_pagetable objects using iommu_domains with IOMMU driver specific parameters. This infrastructure will allow linking those domains to IOAS's and devices. Link: https://lore.kernel.org/r/12-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:35 -04:00
};
iommufd: Add a nested HW pagetable object IOMMU_HWPT_ALLOC already supports iommu_domain allocation for usersapce. But it can only allocate a hw_pagetable that associates to a given IOAS, i.e. only a kernel-managed hw_pagetable of IOMMUFD_OBJ_HWPT_PAGING type. IOMMU drivers can now support user-managed hw_pagetables, for two-stage translation use cases that require user data input from the user space. Add a new IOMMUFD_OBJ_HWPT_NESTED type with its abort/destroy(). Pair it with a new iommufd_hwpt_nested structure and its to_hwpt_nested() helper. Update the to_hwpt_paging() helper, so a NESTED-type hw_pagetable can be handled in the callers, for example iommufd_hw_pagetable_enforce_rr(). Screen the inputs including the parent PAGING-type hw_pagetable that has a need of a new nest_parent flag in the iommufd_hwpt_paging structure. Extend the IOMMU_HWPT_ALLOC ioctl to accept an IOMMU driver specific data input which is tagged by the enum iommu_hwpt_data_type. Also, update the @pt_id to accept hwpt_id too besides an ioas_id. Then, use them to allocate a hw_pagetable of IOMMUFD_OBJ_HWPT_NESTED type using the iommufd_hw_pagetable_alloc_nested() allocator. Link: https://lore.kernel.org/r/20231026043938.63898-8-yi.l.liu@intel.com Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Co-developed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-10-25 21:39:35 -07:00
struct iommufd_hwpt_nested {
struct iommufd_hw_pagetable common;
struct iommufd_hwpt_paging *parent;
struct iommufd_viommu *viommu;
iommufd: Add a nested HW pagetable object IOMMU_HWPT_ALLOC already supports iommu_domain allocation for usersapce. But it can only allocate a hw_pagetable that associates to a given IOAS, i.e. only a kernel-managed hw_pagetable of IOMMUFD_OBJ_HWPT_PAGING type. IOMMU drivers can now support user-managed hw_pagetables, for two-stage translation use cases that require user data input from the user space. Add a new IOMMUFD_OBJ_HWPT_NESTED type with its abort/destroy(). Pair it with a new iommufd_hwpt_nested structure and its to_hwpt_nested() helper. Update the to_hwpt_paging() helper, so a NESTED-type hw_pagetable can be handled in the callers, for example iommufd_hw_pagetable_enforce_rr(). Screen the inputs including the parent PAGING-type hw_pagetable that has a need of a new nest_parent flag in the iommufd_hwpt_paging structure. Extend the IOMMU_HWPT_ALLOC ioctl to accept an IOMMU driver specific data input which is tagged by the enum iommu_hwpt_data_type. Also, update the @pt_id to accept hwpt_id too besides an ioas_id. Then, use them to allocate a hw_pagetable of IOMMUFD_OBJ_HWPT_NESTED type using the iommufd_hw_pagetable_alloc_nested() allocator. Link: https://lore.kernel.org/r/20231026043938.63898-8-yi.l.liu@intel.com Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Co-developed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-10-25 21:39:35 -07:00
};
static inline bool hwpt_is_paging(struct iommufd_hw_pagetable *hwpt)
{
return hwpt->obj.type == IOMMUFD_OBJ_HWPT_PAGING;
}
static inline struct iommufd_hwpt_paging *
to_hwpt_paging(struct iommufd_hw_pagetable *hwpt)
{
return container_of(hwpt, struct iommufd_hwpt_paging, common);
}
iommufd/device: Enforce reserved IOVA also when attached to hwpt_nested Currently, device reserved regions are only enforced when the device is attached to an hwpt_paging. In other words, if the device gets attached to an hwpt_nested directly, the parent hwpt_paging of the hwpt_nested's would not enforce those reserved IOVAs. This works for most of reserved region types, but not for IOMMU_RESV_SW_MSI, which is a unique software defined window, required by a nesting case too to setup an MSI doorbell on the parent stage-2 hwpt/domain. Kevin pointed out in 1 that: 1) there is no usage using up closely the entire IOVA space yet, 2) guest may change the viommu mode to switch between nested and paging then VMM has to take all devices' reserved regions into consideration anyway, when composing the GPA space. So it would be actually convenient for us to also enforce reserved IOVA onto the parent hwpt_paging, when attaching a device to an hwpt_nested. Repurpose the existing attach/replace_paging helpers to attach device's reserved IOVAs exclusively. Add a new find_hwpt_paging helper, which is only used by these reserved IOVA functions, to allow an IOMMUFD_OBJ_HWPT_NESTED hwpt to redirect to its parent hwpt_paging. Return a NULL in these two helpers for any new HWPT type in the future. Link: https://patch.msgid.link/r/20240807003446.3740368-1-nicolinc@nvidia.com Link: https://lore.kernel.org/all/BN9PR11MB5276497781C96415272E6FED8CB12@BN9PR11MB5276.namprd11.prod.outlook.com/ #1 Suggested-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2024-08-06 17:34:46 -07:00
static inline struct iommufd_hwpt_nested *
to_hwpt_nested(struct iommufd_hw_pagetable *hwpt)
{
return container_of(hwpt, struct iommufd_hwpt_nested, common);
}
static inline struct iommufd_hwpt_paging *
find_hwpt_paging(struct iommufd_hw_pagetable *hwpt)
{
switch (hwpt->obj.type) {
case IOMMUFD_OBJ_HWPT_PAGING:
return to_hwpt_paging(hwpt);
case IOMMUFD_OBJ_HWPT_NESTED:
return to_hwpt_nested(hwpt)->parent;
default:
return NULL;
}
}
static inline struct iommufd_hwpt_paging *
iommufd_get_hwpt_paging(struct iommufd_ucmd *ucmd, u32 id)
{
return container_of(iommufd_get_object(ucmd->ictx, id,
IOMMUFD_OBJ_HWPT_PAGING),
struct iommufd_hwpt_paging, common.obj);
}
static inline struct iommufd_hw_pagetable *
iommufd_get_hwpt_nested(struct iommufd_ucmd *ucmd, u32 id)
{
return container_of(iommufd_get_object(ucmd->ictx, id,
IOMMUFD_OBJ_HWPT_NESTED),
struct iommufd_hw_pagetable, obj);
}
int iommufd_hwpt_set_dirty_tracking(struct iommufd_ucmd *ucmd);
int iommufd_hwpt_get_dirty_bitmap(struct iommufd_ucmd *ucmd);
struct iommufd_hwpt_paging *
iommufd_hwpt_paging_alloc(struct iommufd_ctx *ictx, struct iommufd_ioas *ioas,
struct iommufd_device *idev, ioasid_t pasid,
u32 flags, bool immediate_attach,
iommufd: Add a nested HW pagetable object IOMMU_HWPT_ALLOC already supports iommu_domain allocation for usersapce. But it can only allocate a hw_pagetable that associates to a given IOAS, i.e. only a kernel-managed hw_pagetable of IOMMUFD_OBJ_HWPT_PAGING type. IOMMU drivers can now support user-managed hw_pagetables, for two-stage translation use cases that require user data input from the user space. Add a new IOMMUFD_OBJ_HWPT_NESTED type with its abort/destroy(). Pair it with a new iommufd_hwpt_nested structure and its to_hwpt_nested() helper. Update the to_hwpt_paging() helper, so a NESTED-type hw_pagetable can be handled in the callers, for example iommufd_hw_pagetable_enforce_rr(). Screen the inputs including the parent PAGING-type hw_pagetable that has a need of a new nest_parent flag in the iommufd_hwpt_paging structure. Extend the IOMMU_HWPT_ALLOC ioctl to accept an IOMMU driver specific data input which is tagged by the enum iommu_hwpt_data_type. Also, update the @pt_id to accept hwpt_id too besides an ioas_id. Then, use them to allocate a hw_pagetable of IOMMUFD_OBJ_HWPT_NESTED type using the iommufd_hw_pagetable_alloc_nested() allocator. Link: https://lore.kernel.org/r/20231026043938.63898-8-yi.l.liu@intel.com Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Co-developed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-10-25 21:39:35 -07:00
const struct iommu_user_data *user_data);
int iommufd_hw_pagetable_attach(struct iommufd_hw_pagetable *hwpt,
struct iommufd_device *idev, ioasid_t pasid);
struct iommufd_hw_pagetable *
iommufd_hw_pagetable_detach(struct iommufd_device *idev, ioasid_t pasid);
void iommufd_hwpt_paging_destroy(struct iommufd_object *obj);
void iommufd_hwpt_paging_abort(struct iommufd_object *obj);
iommufd: Add a nested HW pagetable object IOMMU_HWPT_ALLOC already supports iommu_domain allocation for usersapce. But it can only allocate a hw_pagetable that associates to a given IOAS, i.e. only a kernel-managed hw_pagetable of IOMMUFD_OBJ_HWPT_PAGING type. IOMMU drivers can now support user-managed hw_pagetables, for two-stage translation use cases that require user data input from the user space. Add a new IOMMUFD_OBJ_HWPT_NESTED type with its abort/destroy(). Pair it with a new iommufd_hwpt_nested structure and its to_hwpt_nested() helper. Update the to_hwpt_paging() helper, so a NESTED-type hw_pagetable can be handled in the callers, for example iommufd_hw_pagetable_enforce_rr(). Screen the inputs including the parent PAGING-type hw_pagetable that has a need of a new nest_parent flag in the iommufd_hwpt_paging structure. Extend the IOMMU_HWPT_ALLOC ioctl to accept an IOMMU driver specific data input which is tagged by the enum iommu_hwpt_data_type. Also, update the @pt_id to accept hwpt_id too besides an ioas_id. Then, use them to allocate a hw_pagetable of IOMMUFD_OBJ_HWPT_NESTED type using the iommufd_hw_pagetable_alloc_nested() allocator. Link: https://lore.kernel.org/r/20231026043938.63898-8-yi.l.liu@intel.com Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Co-developed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-10-25 21:39:35 -07:00
void iommufd_hwpt_nested_destroy(struct iommufd_object *obj);
void iommufd_hwpt_nested_abort(struct iommufd_object *obj);
int iommufd_hwpt_alloc(struct iommufd_ucmd *ucmd);
int iommufd_hwpt_invalidate(struct iommufd_ucmd *ucmd);
iommufd: Add a HW pagetable object The hw_pagetable object exposes the internal struct iommu_domain's to userspace. An iommu_domain is required when any DMA device attaches to an IOAS to control the io page table through the iommu driver. For compatibility with VFIO the hw_pagetable is automatically created when a DMA device is attached to the IOAS. If a compatible iommu_domain already exists then the hw_pagetable associated with it is used for the attachment. In the initial series there is no iommufd uAPI for the hw_pagetable object. The next patch provides driver facing APIs for IO page table attachment that allows drivers to accept either an IOAS or a hw_pagetable ID and for the driver to return the hw_pagetable ID that was auto-selected from an IOAS. The expectation is the driver will provide uAPI through its own FD for attaching its device to iommufd. This allows userspace to learn the mapping of devices to iommu_domains and to override the automatic attachment. The future HW specific interface will allow userspace to create hw_pagetable objects using iommu_domains with IOMMU driver specific parameters. This infrastructure will allow linking those domains to IOAS's and devices. Link: https://lore.kernel.org/r/12-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:35 -04:00
static inline void iommufd_hw_pagetable_put(struct iommufd_ctx *ictx,
struct iommufd_hw_pagetable *hwpt)
{
iommufd: Add a nested HW pagetable object IOMMU_HWPT_ALLOC already supports iommu_domain allocation for usersapce. But it can only allocate a hw_pagetable that associates to a given IOAS, i.e. only a kernel-managed hw_pagetable of IOMMUFD_OBJ_HWPT_PAGING type. IOMMU drivers can now support user-managed hw_pagetables, for two-stage translation use cases that require user data input from the user space. Add a new IOMMUFD_OBJ_HWPT_NESTED type with its abort/destroy(). Pair it with a new iommufd_hwpt_nested structure and its to_hwpt_nested() helper. Update the to_hwpt_paging() helper, so a NESTED-type hw_pagetable can be handled in the callers, for example iommufd_hw_pagetable_enforce_rr(). Screen the inputs including the parent PAGING-type hw_pagetable that has a need of a new nest_parent flag in the iommufd_hwpt_paging structure. Extend the IOMMU_HWPT_ALLOC ioctl to accept an IOMMU driver specific data input which is tagged by the enum iommu_hwpt_data_type. Also, update the @pt_id to accept hwpt_id too besides an ioas_id. Then, use them to allocate a hw_pagetable of IOMMUFD_OBJ_HWPT_NESTED type using the iommufd_hw_pagetable_alloc_nested() allocator. Link: https://lore.kernel.org/r/20231026043938.63898-8-yi.l.liu@intel.com Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Co-developed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-10-25 21:39:35 -07:00
if (hwpt->obj.type == IOMMUFD_OBJ_HWPT_PAGING) {
struct iommufd_hwpt_paging *hwpt_paging = to_hwpt_paging(hwpt);
iommufd: Add a nested HW pagetable object IOMMU_HWPT_ALLOC already supports iommu_domain allocation for usersapce. But it can only allocate a hw_pagetable that associates to a given IOAS, i.e. only a kernel-managed hw_pagetable of IOMMUFD_OBJ_HWPT_PAGING type. IOMMU drivers can now support user-managed hw_pagetables, for two-stage translation use cases that require user data input from the user space. Add a new IOMMUFD_OBJ_HWPT_NESTED type with its abort/destroy(). Pair it with a new iommufd_hwpt_nested structure and its to_hwpt_nested() helper. Update the to_hwpt_paging() helper, so a NESTED-type hw_pagetable can be handled in the callers, for example iommufd_hw_pagetable_enforce_rr(). Screen the inputs including the parent PAGING-type hw_pagetable that has a need of a new nest_parent flag in the iommufd_hwpt_paging structure. Extend the IOMMU_HWPT_ALLOC ioctl to accept an IOMMU driver specific data input which is tagged by the enum iommu_hwpt_data_type. Also, update the @pt_id to accept hwpt_id too besides an ioas_id. Then, use them to allocate a hw_pagetable of IOMMUFD_OBJ_HWPT_NESTED type using the iommufd_hw_pagetable_alloc_nested() allocator. Link: https://lore.kernel.org/r/20231026043938.63898-8-yi.l.liu@intel.com Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Co-developed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-10-25 21:39:35 -07:00
lockdep_assert_not_held(&hwpt_paging->ioas->mutex);
if (hwpt_paging->auto_domain) {
iommufd: Do not UAF during iommufd_put_object() The mixture of kernel and user space lifecycle objects continues to be complicated inside iommufd. The obj->destroy_rwsem is used to bring order to the kernel driver destruction sequence but it cannot be sequenced right with the other refcounts so we end up possibly UAF'ing: BUG: KASAN: slab-use-after-free in __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 Read of size 8 at addr ffff888073cde868 by task syz-executor934/6535 CPU: 1 PID: 6535 Comm: syz-executor934 Not tainted 6.6.0-rc7-syzkaller-00195-g2af9b20dbb39 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/09/2023 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc4/0x620 mm/kasan/report.c:475 kasan_report+0xda/0x110 mm/kasan/report.c:588 __up_read+0x627/0x750 kernel/locking/rwsem.c:1342 iommufd_put_object drivers/iommu/iommufd/iommufd_private.h:149 [inline] iommufd_vfio_ioas+0x46c/0x580 drivers/iommu/iommufd/vfio_compat.c:146 iommufd_fops_ioctl+0x347/0x4d0 drivers/iommu/iommufd/main.c:398 vfs_ioctl fs/ioctl.c:51 [inline] __do_sys_ioctl fs/ioctl.c:871 [inline] __se_sys_ioctl fs/ioctl.c:857 [inline] __x64_sys_ioctl+0x18f/0x210 fs/ioctl.c:857 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd There are two races here, the more obvious one: CPU 0 CPU 1 iommufd_put_object() iommufd_destroy() refcount_dec(&obj->users) iommufd_object_remove() kfree() up_read(&obj->destroy_rwsem) // Boom And there is also perhaps some possibility that the rwsem could hit an issue: CPU 0 CPU 1 iommufd_put_object() iommufd_object_destroy_user() refcount_dec(&obj->users); down_write(&obj->destroy_rwsem) up_read(&obj->destroy_rwsem); atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count); tmp = atomic_long_add_return_release() rwsem_try_write_lock() iommufd_object_remove() up_write(&obj->destroy_rwsem) kfree() clear_nonspinnable() // Boom Fix this by reorganizing this again so that two refcounts are used to keep track of things with a rule that users == 0 && shortterm_users == 0 means no other threads have that memory. Put a wait_queue in the iommufd_ctx object that is triggered when any sub object reaches a 0 shortterm_users. This allows the same wait for userspace ioctls to finish behavior that the rwsem was providing. This is weaker still than the prior versions: - There is no bias on shortterm_users so if some thread is waiting to destroy other threads can continue to get new read sides - If destruction fails, eg because of an active in-kernel user, then shortterm_users will have cycled to zero momentarily blocking new users - If userspace races destroy with other userspace operations they continue to get an EBUSY since we still can't intermix looking up an ID and sleeping for its unref In all cases these are things that userspace brings on itself, correct programs will not hit them. Fixes: 99f98a7c0d69 ("iommufd: IOMMUFD_DESTROY should not increase the refcount") Link: https://lore.kernel.org/all/2-v2-ca9e00171c5b+123-iommufd_syz4_jgg@nvidia.com/ Reported-by: syzbot+d31adfb277377ef8fcba@syzkaller.appspotmail.com Closes: https://lore.kernel.org/r/00000000000055ef9a0609336580@google.com Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-11-12 15:44:08 -04:00
iommufd_object_put_and_try_destroy(ictx, &hwpt->obj);
iommufd: Add a nested HW pagetable object IOMMU_HWPT_ALLOC already supports iommu_domain allocation for usersapce. But it can only allocate a hw_pagetable that associates to a given IOAS, i.e. only a kernel-managed hw_pagetable of IOMMUFD_OBJ_HWPT_PAGING type. IOMMU drivers can now support user-managed hw_pagetables, for two-stage translation use cases that require user data input from the user space. Add a new IOMMUFD_OBJ_HWPT_NESTED type with its abort/destroy(). Pair it with a new iommufd_hwpt_nested structure and its to_hwpt_nested() helper. Update the to_hwpt_paging() helper, so a NESTED-type hw_pagetable can be handled in the callers, for example iommufd_hw_pagetable_enforce_rr(). Screen the inputs including the parent PAGING-type hw_pagetable that has a need of a new nest_parent flag in the iommufd_hwpt_paging structure. Extend the IOMMU_HWPT_ALLOC ioctl to accept an IOMMU driver specific data input which is tagged by the enum iommu_hwpt_data_type. Also, update the @pt_id to accept hwpt_id too besides an ioas_id. Then, use them to allocate a hw_pagetable of IOMMUFD_OBJ_HWPT_NESTED type using the iommufd_hw_pagetable_alloc_nested() allocator. Link: https://lore.kernel.org/r/20231026043938.63898-8-yi.l.liu@intel.com Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Co-developed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2023-10-25 21:39:35 -07:00
return;
}
}
refcount_dec(&hwpt->obj.users);
}
struct iommufd_attach;
struct iommufd_group {
struct kref ref;
struct mutex lock;
struct iommufd_ctx *ictx;
struct iommu_group *group;
struct xarray pasid_attach;
struct iommufd_sw_msi_maps required_sw_msi;
phys_addr_t sw_msi_start;
};
/*
* A iommufd_device object represents the binding relationship between a
* consuming driver and the iommufd. These objects are created/destroyed by
* external drivers, not by userspace.
*/
struct iommufd_device {
struct iommufd_object obj;
struct iommufd_ctx *ictx;
struct iommufd_group *igroup;
struct list_head group_item;
/* always the physical device */
struct device *dev;
bool enforce_cache_coherency;
iommufd: Destroy vdevice on idevice destroy Destroy iommufd_vdevice (vdev) on iommufd_idevice (idev) destruction so that vdev can't outlive idev. idev represents the physical device bound to iommufd, while the vdev represents the virtual instance of the physical device in the VM. The lifecycle of the vdev should not be longer than idev. This doesn't cause real problem on existing use cases cause vdev doesn't impact the physical device, only provides virtualization information. But to extend vdev for Confidential Computing (CC), there are needs to do secure configuration for the vdev, e.g. TSM Bind/Unbind. These configurations should be rolled back on idev destroy, or the external driver (VFIO) functionality may be impact. The idev is created by external driver so its destruction can't fail. The idev implements pre_destroy() op to actively remove its associated vdev before destroying itself. There are 3 cases on idev pre_destroy(): 1. vdev is already destroyed by userspace. No extra handling needed. 2. vdev is still alive. Use iommufd_object_tombstone_user() to destroy vdev and tombstone the vdev ID. 3. vdev is being destroyed by userspace. The vdev ID is already freed, but vdev destroy handler is not completed. This requires multi-threads syncing - vdev holds idev's short term users reference until vdev destruction completes, idev leverages existing wait_shortterm mechanism for syncing. idev should also block any new reference to it after pre_destroy(), or the following wait shortterm would timeout. Introduce a 'destroying' flag, set it to true on idev pre_destroy(). Any attempt to reference idev should honor this flag under the protection of idev->igroup->lock. Link: https://patch.msgid.link/r/20250716070349.1807226-5-yilun.xu@linux.intel.com Originally-by: Nicolin Chen <nicolinc@nvidia.com> Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Nicolin Chen <nicolinc@nvidia.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Co-developed-by: "Aneesh Kumar K.V (Arm)" <aneesh.kumar@kernel.org> Signed-off-by: "Aneesh Kumar K.V (Arm)" <aneesh.kumar@kernel.org> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Xu Yilun <yilun.xu@linux.intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2025-07-16 15:03:45 +08:00
struct iommufd_vdevice *vdev;
bool destroying;
};
static inline struct iommufd_device *
iommufd_get_device(struct iommufd_ucmd *ucmd, u32 id)
{
return container_of(iommufd_get_object(ucmd->ictx, id,
IOMMUFD_OBJ_DEVICE),
struct iommufd_device, obj);
}
iommufd: Destroy vdevice on idevice destroy Destroy iommufd_vdevice (vdev) on iommufd_idevice (idev) destruction so that vdev can't outlive idev. idev represents the physical device bound to iommufd, while the vdev represents the virtual instance of the physical device in the VM. The lifecycle of the vdev should not be longer than idev. This doesn't cause real problem on existing use cases cause vdev doesn't impact the physical device, only provides virtualization information. But to extend vdev for Confidential Computing (CC), there are needs to do secure configuration for the vdev, e.g. TSM Bind/Unbind. These configurations should be rolled back on idev destroy, or the external driver (VFIO) functionality may be impact. The idev is created by external driver so its destruction can't fail. The idev implements pre_destroy() op to actively remove its associated vdev before destroying itself. There are 3 cases on idev pre_destroy(): 1. vdev is already destroyed by userspace. No extra handling needed. 2. vdev is still alive. Use iommufd_object_tombstone_user() to destroy vdev and tombstone the vdev ID. 3. vdev is being destroyed by userspace. The vdev ID is already freed, but vdev destroy handler is not completed. This requires multi-threads syncing - vdev holds idev's short term users reference until vdev destruction completes, idev leverages existing wait_shortterm mechanism for syncing. idev should also block any new reference to it after pre_destroy(), or the following wait shortterm would timeout. Introduce a 'destroying' flag, set it to true on idev pre_destroy(). Any attempt to reference idev should honor this flag under the protection of idev->igroup->lock. Link: https://patch.msgid.link/r/20250716070349.1807226-5-yilun.xu@linux.intel.com Originally-by: Nicolin Chen <nicolinc@nvidia.com> Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Nicolin Chen <nicolinc@nvidia.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Co-developed-by: "Aneesh Kumar K.V (Arm)" <aneesh.kumar@kernel.org> Signed-off-by: "Aneesh Kumar K.V (Arm)" <aneesh.kumar@kernel.org> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Xu Yilun <yilun.xu@linux.intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2025-07-16 15:03:45 +08:00
void iommufd_device_pre_destroy(struct iommufd_object *obj);
void iommufd_device_destroy(struct iommufd_object *obj);
int iommufd_get_hw_info(struct iommufd_ucmd *ucmd);
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
struct iommufd_access {
struct iommufd_object obj;
struct iommufd_ctx *ictx;
struct iommufd_ioas *ioas;
iommufd/device: Add iommufd_access_detach() API Previously, the detach routine is only done by the destroy(). And it was called by vfio_iommufd_emulated_unbind() when the device runs close(), so all the mappings in iopt were cleaned in that setup, when the call trace reaches this detach() routine. Now, there's a need of a detach uAPI, meaning that it does not only need a new iommufd_access_detach() API, but also requires access->ops->unmap() call as a cleanup. So add one. However, leaving that unprotected can introduce some potential of a race condition during the pin_/unpin_pages() call, where access->ioas->iopt is getting referenced. So, add an ioas_lock to protect the context of iopt referencings. Also, to allow the iommufd_access_unpin_pages() callback to happen via this unmap() call, add an ioas_unpin pointer, so the unpin routine won't be affected by the "access->ioas = NULL" trick. Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Tested-by: Terrence Xu <terrence.xu@intel.com> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Tested-by: Yanting Jiang <yanting.jiang@intel.com> Tested-by: Shameer Kolothum <shameerali.kolothum.thodi@huawei.com> Tested-by: Zhenzhong Duan <zhenzhong.duan@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Link: https://lore.kernel.org/r/20230718135551.6592-15-yi.l.liu@intel.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2023-07-18 06:55:39 -07:00
struct iommufd_ioas *ioas_unpin;
struct mutex ioas_lock;
const struct iommufd_access_ops *ops;
void *data;
iommufd: Data structure to provide IOVA to PFN mapping This is the remainder of the IOAS data structure. Provide an object called an io_pagetable that is composed of iopt_areas pointing at iopt_pages, along with a list of iommu_domains that mirror the IOVA to PFN map. At the top this is a simple interval tree of iopt_areas indicating the map of IOVA to iopt_pages. An xarray keeps track of a list of domains. Based on the attached domains there is a minimum alignment for areas (which may be smaller than PAGE_SIZE), an interval tree of reserved IOVA that can't be mapped and an IOVA of allowed IOVA that can always be mappable. The concept of an 'access' refers to something like a VFIO mdev that is accessing the IOVA and using a 'struct page *' for CPU based access. Externally an API is provided that matches the requirements of the IOCTL interface for map/unmap and domain attachment. The API provides a 'copy' primitive to establish a new IOVA map in a different IOAS from an existing mapping by re-using the iopt_pages. This is the basic mechanism to provide single pinning. This is designed to support a pre-registration flow where userspace would setup an dummy IOAS with no domains, map in memory and then establish an access to pin all PFNs into the xarray. Copy can then be used to create new IOVA mappings in a different IOAS, with iommu_domains attached. Upon copy the PFNs will be read out of the xarray and mapped into the iommu_domains, avoiding any pin_user_pages() overheads. Link: https://lore.kernel.org/r/10-v6-a196d26f289e+11787-iommufd_jgg@nvidia.com Tested-by: Nicolin Chen <nicolinc@nvidia.com> Tested-by: Yi Liu <yi.l.liu@intel.com> Tested-by: Lixiao Yang <lixiao.yang@intel.com> Tested-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2022-11-29 16:29:33 -04:00
unsigned long iova_alignment;
u32 iopt_access_list_id;
};
int iopt_add_access(struct io_pagetable *iopt, struct iommufd_access *access);
void iopt_remove_access(struct io_pagetable *iopt,
struct iommufd_access *access, u32 iopt_access_list_id);
void iommufd_access_destroy_object(struct iommufd_object *obj);
/* iommufd_access for internal use */
static inline bool iommufd_access_is_internal(struct iommufd_access *access)
{
return !access->ictx;
}
struct iommufd_access *iommufd_access_create_internal(struct iommufd_ctx *ictx);
static inline void
iommufd_access_destroy_internal(struct iommufd_ctx *ictx,
struct iommufd_access *access)
{
iommufd_object_destroy_user(ictx, &access->obj);
}
int iommufd_access_attach_internal(struct iommufd_access *access,
struct iommufd_ioas *ioas);
static inline void iommufd_access_detach_internal(struct iommufd_access *access)
{
iommufd_access_detach(access);
}
struct iommufd_eventq {
struct iommufd_object obj;
struct iommufd_ctx *ictx;
struct file *filep;
iommufd/fault: Use a separate spinlock to protect fault->deliver list The fault->mutex serializes the fault read()/write() fops and the iommufd_fault_auto_response_faults(), mainly for fault->response. Also, it was conveniently used to fence the fault->deliver in poll() fop and iommufd_fault_iopf_handler(). However, copy_from/to_user() may sleep if pagefaults are enabled. Thus, they could take a long time to wait for user pages to swap in, blocking iommufd_fault_iopf_handler() and its caller that is typically a shared IRQ handler of an IOMMU driver, resulting in a potential global DOS. Instead of reusing the mutex to protect the fault->deliver list, add a separate spinlock, nested under the mutex, to do the job. iommufd_fault_iopf_handler() would no longer be blocked by copy_from/to_user(). Add a free_list in iommufd_auto_response_faults(), so the spinlock can simply fence a fast list_for_each_entry_safe routine. Provide two deliver list helpers for iommufd_fault_fops_read() to use: - Fetch the first iopf_group out of the fault->deliver list - Restore an iopf_group back to the head of the fault->deliver list Lastly, move the mutex closer to the response in the fault structure, and update its kdoc accordingly. Fixes: 07838f7fd529 ("iommufd: Add iommufd fault object") Link: https://patch.msgid.link/r/20250117192901.79491-1-nicolinc@nvidia.com Cc: stable@vger.kernel.org Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2025-01-17 11:29:01 -08:00
spinlock_t lock; /* protects the deliver list */
struct list_head deliver;
struct wait_queue_head wait_queue;
};
struct iommufd_attach_handle {
struct iommu_attach_handle handle;
struct iommufd_device *idev;
};
/* Convert an iommu attach handle to iommufd handle. */
#define to_iommufd_handle(hdl) container_of(hdl, struct iommufd_attach_handle, handle)
/*
* An iommufd_fault object represents an interface to deliver I/O page faults
* to the user space. These objects are created/destroyed by the user space and
* associated with hardware page table objects during page-table allocation.
*/
struct iommufd_fault {
struct iommufd_eventq common;
struct mutex mutex; /* serializes response flows */
struct xarray response;
};
static inline struct iommufd_fault *
eventq_to_fault(struct iommufd_eventq *eventq)
{
return container_of(eventq, struct iommufd_fault, common);
}
static inline struct iommufd_fault *
iommufd_get_fault(struct iommufd_ucmd *ucmd, u32 id)
{
return container_of(iommufd_get_object(ucmd->ictx, id,
IOMMUFD_OBJ_FAULT),
struct iommufd_fault, common.obj);
}
int iommufd_fault_alloc(struct iommufd_ucmd *ucmd);
void iommufd_fault_destroy(struct iommufd_object *obj);
int iommufd_fault_iopf_handler(struct iopf_group *group);
void iommufd_auto_response_faults(struct iommufd_hw_pagetable *hwpt,
struct iommufd_attach_handle *handle);
/* An iommufd_vevent represents a vIOMMU event in an iommufd_veventq */
struct iommufd_vevent {
struct iommufd_vevent_header header;
struct list_head node; /* for iommufd_eventq::deliver */
ssize_t data_len;
u64 event_data[] __counted_by(data_len);
};
#define vevent_for_lost_events_header(vevent) \
(vevent->header.flags & IOMMU_VEVENTQ_FLAG_LOST_EVENTS)
/*
* An iommufd_veventq object represents an interface to deliver vIOMMU events to
* the user space. It is created/destroyed by the user space and associated with
* a vIOMMU object during the allocations.
*/
struct iommufd_veventq {
struct iommufd_eventq common;
struct iommufd_viommu *viommu;
struct list_head node; /* for iommufd_viommu::veventqs */
struct iommufd_vevent lost_events_header;
enum iommu_veventq_type type;
unsigned int depth;
/* Use common.lock for protection */
u32 num_events;
u32 sequence;
};
static inline struct iommufd_veventq *
eventq_to_veventq(struct iommufd_eventq *eventq)
{
return container_of(eventq, struct iommufd_veventq, common);
}
static inline struct iommufd_veventq *
iommufd_get_veventq(struct iommufd_ucmd *ucmd, u32 id)
{
return container_of(iommufd_get_object(ucmd->ictx, id,
IOMMUFD_OBJ_VEVENTQ),
struct iommufd_veventq, common.obj);
}
int iommufd_veventq_alloc(struct iommufd_ucmd *ucmd);
void iommufd_veventq_destroy(struct iommufd_object *obj);
void iommufd_veventq_abort(struct iommufd_object *obj);
static inline void iommufd_vevent_handler(struct iommufd_veventq *veventq,
struct iommufd_vevent *vevent)
{
struct iommufd_eventq *eventq = &veventq->common;
lockdep_assert_held(&eventq->lock);
/*
* Remove the lost_events_header and add the new node at the same time.
* Note the new node can be lost_events_header, for a sequence update.
*/
if (list_is_last(&veventq->lost_events_header.node, &eventq->deliver))
list_del(&veventq->lost_events_header.node);
list_add_tail(&vevent->node, &eventq->deliver);
vevent->header.sequence = veventq->sequence;
veventq->sequence = (veventq->sequence + 1) & INT_MAX;
wake_up_interruptible(&eventq->wait_queue);
}
static inline struct iommufd_viommu *
iommufd_get_viommu(struct iommufd_ucmd *ucmd, u32 id)
{
return container_of(iommufd_get_object(ucmd->ictx, id,
IOMMUFD_OBJ_VIOMMU),
struct iommufd_viommu, obj);
}
static inline struct iommufd_veventq *
iommufd_viommu_find_veventq(struct iommufd_viommu *viommu,
enum iommu_veventq_type type)
{
struct iommufd_veventq *veventq, *next;
lockdep_assert_held(&viommu->veventqs_rwsem);
list_for_each_entry_safe(veventq, next, &viommu->veventqs, node) {
if (veventq->type == type)
return veventq;
}
return NULL;
}
int iommufd_viommu_alloc_ioctl(struct iommufd_ucmd *ucmd);
void iommufd_viommu_destroy(struct iommufd_object *obj);
int iommufd_vdevice_alloc_ioctl(struct iommufd_ucmd *ucmd);
void iommufd_vdevice_destroy(struct iommufd_object *obj);
iommufd: Destroy vdevice on idevice destroy Destroy iommufd_vdevice (vdev) on iommufd_idevice (idev) destruction so that vdev can't outlive idev. idev represents the physical device bound to iommufd, while the vdev represents the virtual instance of the physical device in the VM. The lifecycle of the vdev should not be longer than idev. This doesn't cause real problem on existing use cases cause vdev doesn't impact the physical device, only provides virtualization information. But to extend vdev for Confidential Computing (CC), there are needs to do secure configuration for the vdev, e.g. TSM Bind/Unbind. These configurations should be rolled back on idev destroy, or the external driver (VFIO) functionality may be impact. The idev is created by external driver so its destruction can't fail. The idev implements pre_destroy() op to actively remove its associated vdev before destroying itself. There are 3 cases on idev pre_destroy(): 1. vdev is already destroyed by userspace. No extra handling needed. 2. vdev is still alive. Use iommufd_object_tombstone_user() to destroy vdev and tombstone the vdev ID. 3. vdev is being destroyed by userspace. The vdev ID is already freed, but vdev destroy handler is not completed. This requires multi-threads syncing - vdev holds idev's short term users reference until vdev destruction completes, idev leverages existing wait_shortterm mechanism for syncing. idev should also block any new reference to it after pre_destroy(), or the following wait shortterm would timeout. Introduce a 'destroying' flag, set it to true on idev pre_destroy(). Any attempt to reference idev should honor this flag under the protection of idev->igroup->lock. Link: https://patch.msgid.link/r/20250716070349.1807226-5-yilun.xu@linux.intel.com Originally-by: Nicolin Chen <nicolinc@nvidia.com> Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Nicolin Chen <nicolinc@nvidia.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Co-developed-by: "Aneesh Kumar K.V (Arm)" <aneesh.kumar@kernel.org> Signed-off-by: "Aneesh Kumar K.V (Arm)" <aneesh.kumar@kernel.org> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Xu Yilun <yilun.xu@linux.intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2025-07-16 15:03:45 +08:00
void iommufd_vdevice_abort(struct iommufd_object *obj);
int iommufd_hw_queue_alloc_ioctl(struct iommufd_ucmd *ucmd);
void iommufd_hw_queue_destroy(struct iommufd_object *obj);
iommufd: Destroy vdevice on idevice destroy Destroy iommufd_vdevice (vdev) on iommufd_idevice (idev) destruction so that vdev can't outlive idev. idev represents the physical device bound to iommufd, while the vdev represents the virtual instance of the physical device in the VM. The lifecycle of the vdev should not be longer than idev. This doesn't cause real problem on existing use cases cause vdev doesn't impact the physical device, only provides virtualization information. But to extend vdev for Confidential Computing (CC), there are needs to do secure configuration for the vdev, e.g. TSM Bind/Unbind. These configurations should be rolled back on idev destroy, or the external driver (VFIO) functionality may be impact. The idev is created by external driver so its destruction can't fail. The idev implements pre_destroy() op to actively remove its associated vdev before destroying itself. There are 3 cases on idev pre_destroy(): 1. vdev is already destroyed by userspace. No extra handling needed. 2. vdev is still alive. Use iommufd_object_tombstone_user() to destroy vdev and tombstone the vdev ID. 3. vdev is being destroyed by userspace. The vdev ID is already freed, but vdev destroy handler is not completed. This requires multi-threads syncing - vdev holds idev's short term users reference until vdev destruction completes, idev leverages existing wait_shortterm mechanism for syncing. idev should also block any new reference to it after pre_destroy(), or the following wait shortterm would timeout. Introduce a 'destroying' flag, set it to true on idev pre_destroy(). Any attempt to reference idev should honor this flag under the protection of idev->igroup->lock. Link: https://patch.msgid.link/r/20250716070349.1807226-5-yilun.xu@linux.intel.com Originally-by: Nicolin Chen <nicolinc@nvidia.com> Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Nicolin Chen <nicolinc@nvidia.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Co-developed-by: "Aneesh Kumar K.V (Arm)" <aneesh.kumar@kernel.org> Signed-off-by: "Aneesh Kumar K.V (Arm)" <aneesh.kumar@kernel.org> Tested-by: Nicolin Chen <nicolinc@nvidia.com> Signed-off-by: Xu Yilun <yilun.xu@linux.intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2025-07-16 15:03:45 +08:00
static inline struct iommufd_vdevice *
iommufd_get_vdevice(struct iommufd_ctx *ictx, u32 id)
{
return container_of(iommufd_get_object(ictx, id,
IOMMUFD_OBJ_VDEVICE),
struct iommufd_vdevice, obj);
}
#ifdef CONFIG_IOMMUFD_TEST
int iommufd_test(struct iommufd_ucmd *ucmd);
void iommufd_selftest_destroy(struct iommufd_object *obj);
extern size_t iommufd_test_memory_limit;
void iommufd_test_syz_conv_iova_id(struct iommufd_ucmd *ucmd,
unsigned int ioas_id, u64 *iova, u32 *flags);
bool iommufd_should_fail(void);
int __init iommufd_test_init(void);
void iommufd_test_exit(void);
bool iommufd_selftest_is_mock_dev(struct device *dev);
#else
static inline void iommufd_test_syz_conv_iova_id(struct iommufd_ucmd *ucmd,
unsigned int ioas_id,
u64 *iova, u32 *flags)
{
}
static inline bool iommufd_should_fail(void)
{
return false;
}
static inline int __init iommufd_test_init(void)
{
return 0;
}
static inline void iommufd_test_exit(void)
{
}
static inline bool iommufd_selftest_is_mock_dev(struct device *dev)
{
return false;
}
#endif
#endif