November 2020 - Linux-stable-mirror

FAILED: patch "[PATCH] IB/hfi1: Ensure correct mm is used at all times" failed to apply to 4.9-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.9-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 3d2a9d642512c21a12d19b9250e7a835dcb41a79 Mon Sep 17 00:00:00 2001 From: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> Date: Wed, 25 Nov 2020 16:01:12 -0500 Subject: [PATCH] IB/hfi1: Ensure correct mm is used at all times Two earlier bug fixes have created a security problem in the hfi1 driver. One fix aimed to solve an issue where current->mm was not valid when closing the hfi1 cdev. It attempted to do this by saving a cached value of the current->mm pointer at file open time. This is a problem if another process with access to the FD calls in via write() or ioctl() to pin pages via the hfi driver. The other fix tried to solve a use after free by taking a reference on the mm. To fix this correctly we use the existing cached value of the mm in the mmu notifier. Now we can check in the insert, evict, etc. routines that current->mm matched what the notifier was registered for. If not, then don't allow access. The register of the mmu notifier will save the mm pointer. Since in do_exit() the exit_mm() is called before exit_files(), which would call our close routine a reference is needed on the mm. We rely on the mmgrab done by the registration of the notifier, whereas before it was explicit. The mmu notifier deregistration happens when the user context is torn down, the creation of which triggered the registration. Also of note is we do not do any explicit work to protect the interval tree notifier. It doesn't seem that this is going to be needed since we aren't actually doing anything with current->mm. The interval tree notifier stuff still has a FIXME noted from a previous commit that will be addressed in a follow on patch. Cc: <stable(a)vger.kernel.org> Fixes: e0cf75deab81 ("IB/hfi1: Fix mm_struct use after free") Fixes: 3faa3d9a308e ("IB/hfi1: Make use of mm consistent") Link: https://lore.kernel.org/r/20201125210112.104301.51331.stgit@awfm-01.aw.inte… Suggested-by: Jann Horn <jannh(a)google.com> Reported-by: Jason Gunthorpe <jgg(a)nvidia.com> Reviewed-by: Ira Weiny <ira.weiny(a)intel.com> Reviewed-by: Mike Marciniszyn <mike.marciniszyn(a)cornelisnetworks.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> Signed-off-by: Jason Gunthorpe <jgg(a)nvidia.com> diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c index 8ca51e43cf53..329ee4f48d95 100644 --- a/drivers/infiniband/hw/hfi1/file_ops.c +++ b/drivers/infiniband/hw/hfi1/file_ops.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2020 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -206,8 +207,6 @@ static int hfi1_file_open(struct inode *inode, struct file *fp) spin_lock_init(&fd->tid_lock); spin_lock_init(&fd->invalid_lock); fd->rec_cpu_num = -1; /* no cpu affinity by default */ - fd->mm = current->mm; - mmgrab(fd->mm); fd->dd = dd; fp->private_data = fd; return 0; @@ -711,7 +710,6 @@ static int hfi1_file_close(struct inode *inode, struct file *fp) deallocate_ctxt(uctxt); done: - mmdrop(fdata->mm); if (atomic_dec_and_test(&dd->user_refcount)) complete(&dd->user_comp); diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h index b4c6bff60a4e..e09e8244a94c 100644 --- a/drivers/infiniband/hw/hfi1/hfi.h +++ b/drivers/infiniband/hw/hfi1/hfi.h @@ -1,6 +1,7 @@ #ifndef _HFI1_KERNEL_H #define _HFI1_KERNEL_H /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2020 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -1451,7 +1452,6 @@ struct hfi1_filedata { u32 invalid_tid_idx; /* protect invalid_tids array and invalid_tid_idx */ spinlock_t invalid_lock; - struct mm_struct *mm; }; extern struct xarray hfi1_dev_table; diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.c b/drivers/infiniband/hw/hfi1/mmu_rb.c index 24ca17b77b72..f3fb28e3d5d7 100644 --- a/drivers/infiniband/hw/hfi1/mmu_rb.c +++ b/drivers/infiniband/hw/hfi1/mmu_rb.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2016 - 2017 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -48,23 +49,11 @@ #include <linux/rculist.h> #include <linux/mmu_notifier.h> #include <linux/interval_tree_generic.h> +#include <linux/sched/mm.h> #include "mmu_rb.h" #include "trace.h" -struct mmu_rb_handler { - struct mmu_notifier mn; - struct rb_root_cached root; - void *ops_arg; - spinlock_t lock; /* protect the RB tree */ - struct mmu_rb_ops *ops; - struct mm_struct *mm; - struct list_head lru_list; - struct work_struct del_work; - struct list_head del_list; - struct workqueue_struct *wq; -}; - static unsigned long mmu_node_start(struct mmu_rb_node *); static unsigned long mmu_node_last(struct mmu_rb_node *); static int mmu_notifier_range_start(struct mmu_notifier *, @@ -92,37 +81,36 @@ static unsigned long mmu_node_last(struct mmu_rb_node *node) return PAGE_ALIGN(node->addr + node->len) - 1; } -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm, +int hfi1_mmu_rb_register(void *ops_arg, struct mmu_rb_ops *ops, struct workqueue_struct *wq, struct mmu_rb_handler **handler) { - struct mmu_rb_handler *handlr; + struct mmu_rb_handler *h; int ret; - handlr = kmalloc(sizeof(*handlr), GFP_KERNEL); - if (!handlr) + h = kmalloc(sizeof(*h), GFP_KERNEL); + if (!h) return -ENOMEM; - handlr->root = RB_ROOT_CACHED; - handlr->ops = ops; - handlr->ops_arg = ops_arg; - INIT_HLIST_NODE(&handlr->mn.hlist); - spin_lock_init(&handlr->lock); - handlr->mn.ops = &mn_opts; - handlr->mm = mm; - INIT_WORK(&handlr->del_work, handle_remove); - INIT_LIST_HEAD(&handlr->del_list); - INIT_LIST_HEAD(&handlr->lru_list); - handlr->wq = wq; - - ret = mmu_notifier_register(&handlr->mn, handlr->mm); + h->root = RB_ROOT_CACHED; + h->ops = ops; + h->ops_arg = ops_arg; + INIT_HLIST_NODE(&h->mn.hlist); + spin_lock_init(&h->lock); + h->mn.ops = &mn_opts; + INIT_WORK(&h->del_work, handle_remove); + INIT_LIST_HEAD(&h->del_list); + INIT_LIST_HEAD(&h->lru_list); + h->wq = wq; + + ret = mmu_notifier_register(&h->mn, current->mm); if (ret) { - kfree(handlr); + kfree(h); return ret; } - *handler = handlr; + *handler = h; return 0; } @@ -134,7 +122,7 @@ void hfi1_mmu_rb_unregister(struct mmu_rb_handler *handler) struct list_head del_list; /* Unregister first so we don't get any more notifications. */ - mmu_notifier_unregister(&handler->mn, handler->mm); + mmu_notifier_unregister(&handler->mn, handler->mn.mm); /* * Make sure the wq delete handler is finished running. It will not @@ -166,6 +154,10 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler, int ret = 0; trace_hfi1_mmu_rb_insert(mnode->addr, mnode->len); + + if (current->mm != handler->mn.mm) + return -EPERM; + spin_lock_irqsave(&handler->lock, flags); node = __mmu_rb_search(handler, mnode->addr, mnode->len); if (node) { @@ -180,6 +172,7 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler, __mmu_int_rb_remove(mnode, &handler->root); list_del(&mnode->list); /* remove from LRU list */ } + mnode->handler = handler; unlock: spin_unlock_irqrestore(&handler->lock, flags); return ret; @@ -217,6 +210,9 @@ bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler, unsigned long flags; bool ret = false; + if (current->mm != handler->mn.mm) + return ret; + spin_lock_irqsave(&handler->lock, flags); node = __mmu_rb_search(handler, addr, len); if (node) { @@ -239,6 +235,9 @@ void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg) unsigned long flags; bool stop = false; + if (current->mm != handler->mn.mm) + return; + INIT_LIST_HEAD(&del_list); spin_lock_irqsave(&handler->lock, flags); @@ -272,6 +271,9 @@ void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler, { unsigned long flags; + if (current->mm != handler->mn.mm) + return; + /* Validity of handler and node pointers has been checked by caller. */ trace_hfi1_mmu_rb_remove(node->addr, node->len); spin_lock_irqsave(&handler->lock, flags); diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.h b/drivers/infiniband/hw/hfi1/mmu_rb.h index f04cec1e99d1..423aacc67e94 100644 --- a/drivers/infiniband/hw/hfi1/mmu_rb.h +++ b/drivers/infiniband/hw/hfi1/mmu_rb.h @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2016 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -54,6 +55,7 @@ struct mmu_rb_node { unsigned long len; unsigned long __last; struct rb_node node; + struct mmu_rb_handler *handler; struct list_head list; }; @@ -71,7 +73,19 @@ struct mmu_rb_ops { void *evict_arg, bool *stop); }; -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm, +struct mmu_rb_handler { + struct mmu_notifier mn; + struct rb_root_cached root; + void *ops_arg; + spinlock_t lock; /* protect the RB tree */ + struct mmu_rb_ops *ops; + struct list_head lru_list; + struct work_struct del_work; + struct list_head del_list; + struct workqueue_struct *wq; +}; + +int hfi1_mmu_rb_register(void *ops_arg, struct mmu_rb_ops *ops, struct workqueue_struct *wq, struct mmu_rb_handler **handler); diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.c b/drivers/infiniband/hw/hfi1/user_exp_rcv.c index f81ca20f4b69..b94fc7fd75a9 100644 --- a/drivers/infiniband/hw/hfi1/user_exp_rcv.c +++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -173,15 +174,18 @@ static void unpin_rcv_pages(struct hfi1_filedata *fd, { struct page **pages; struct hfi1_devdata *dd = fd->uctxt->dd; + struct mm_struct *mm; if (mapped) { pci_unmap_single(dd->pcidev, node->dma_addr, node->npages * PAGE_SIZE, PCI_DMA_FROMDEVICE); pages = &node->pages[idx]; + mm = mm_from_tid_node(node); } else { pages = &tidbuf->pages[idx]; + mm = current->mm; } - hfi1_release_user_pages(fd->mm, pages, npages, mapped); + hfi1_release_user_pages(mm, pages, npages, mapped); fd->tid_n_pinned -= npages; } @@ -216,12 +220,12 @@ static int pin_rcv_pages(struct hfi1_filedata *fd, struct tid_user_buf *tidbuf) * pages, accept the amount pinned so far and program only that. * User space knows how to deal with partially programmed buffers. */ - if (!hfi1_can_pin_pages(dd, fd->mm, fd->tid_n_pinned, npages)) { + if (!hfi1_can_pin_pages(dd, current->mm, fd->tid_n_pinned, npages)) { kfree(pages); return -ENOMEM; } - pinned = hfi1_acquire_user_pages(fd->mm, vaddr, npages, true, pages); + pinned = hfi1_acquire_user_pages(current->mm, vaddr, npages, true, pages); if (pinned <= 0) { kfree(pages); return pinned; @@ -756,7 +760,7 @@ static int set_rcvarray_entry(struct hfi1_filedata *fd, if (fd->use_mn) { ret = mmu_interval_notifier_insert( - &node->notifier, fd->mm, + &node->notifier, current->mm, tbuf->vaddr + (pageidx * PAGE_SIZE), npages * PAGE_SIZE, &tid_mn_ops); if (ret) diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.h b/drivers/infiniband/hw/hfi1/user_exp_rcv.h index 332abb446861..d45c7b6988d4 100644 --- a/drivers/infiniband/hw/hfi1/user_exp_rcv.h +++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.h @@ -1,6 +1,7 @@ #ifndef _HFI1_USER_EXP_RCV_H #define _HFI1_USER_EXP_RCV_H /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2017 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -95,4 +96,9 @@ int hfi1_user_exp_rcv_clear(struct hfi1_filedata *fd, int hfi1_user_exp_rcv_invalid(struct hfi1_filedata *fd, struct hfi1_tid_info *tinfo); +static inline struct mm_struct *mm_from_tid_node(struct tid_rb_node *node) +{ + return node->notifier.mm; +} + #endif /* _HFI1_USER_EXP_RCV_H */ diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c index a92346e88628..4a4956f96a7e 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.c +++ b/drivers/infiniband/hw/hfi1/user_sdma.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -188,7 +189,6 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, atomic_set(&pq->n_reqs, 0); init_waitqueue_head(&pq->wait); atomic_set(&pq->n_locked, 0); - pq->mm = fd->mm; iowait_init(&pq->busy, 0, NULL, NULL, defer_packet_queue, activate_packet_queue, NULL, NULL); @@ -230,7 +230,7 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, cq->nentries = hfi1_sdma_comp_ring_size; - ret = hfi1_mmu_rb_register(pq, pq->mm, &sdma_rb_ops, dd->pport->hfi1_wq, + ret = hfi1_mmu_rb_register(pq, &sdma_rb_ops, dd->pport->hfi1_wq, &pq->handler); if (ret) { dd_dev_err(dd, "Failed to register with MMU %d", ret); @@ -980,13 +980,13 @@ static int pin_sdma_pages(struct user_sdma_request *req, npages -= node->npages; retry: - if (!hfi1_can_pin_pages(pq->dd, pq->mm, + if (!hfi1_can_pin_pages(pq->dd, current->mm, atomic_read(&pq->n_locked), npages)) { cleared = sdma_cache_evict(pq, npages); if (cleared >= npages) goto retry; } - pinned = hfi1_acquire_user_pages(pq->mm, + pinned = hfi1_acquire_user_pages(current->mm, ((unsigned long)iovec->iov.iov_base + (node->npages * PAGE_SIZE)), npages, 0, pages + node->npages); @@ -995,7 +995,7 @@ static int pin_sdma_pages(struct user_sdma_request *req, return pinned; } if (pinned != npages) { - unpin_vector_pages(pq->mm, pages, node->npages, pinned); + unpin_vector_pages(current->mm, pages, node->npages, pinned); return -EFAULT; } kfree(node->pages); @@ -1008,7 +1008,8 @@ static int pin_sdma_pages(struct user_sdma_request *req, static void unpin_sdma_pages(struct sdma_mmu_node *node) { if (node->npages) { - unpin_vector_pages(node->pq->mm, node->pages, 0, node->npages); + unpin_vector_pages(mm_from_sdma_node(node), node->pages, 0, + node->npages); atomic_sub(node->npages, &node->pq->n_locked); } } diff --git a/drivers/infiniband/hw/hfi1/user_sdma.h b/drivers/infiniband/hw/hfi1/user_sdma.h index 9972e0e6545e..1e8c02fe8ad1 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.h +++ b/drivers/infiniband/hw/hfi1/user_sdma.h @@ -1,6 +1,7 @@ #ifndef _HFI1_USER_SDMA_H #define _HFI1_USER_SDMA_H /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -133,7 +134,6 @@ struct hfi1_user_sdma_pkt_q { unsigned long unpinned; struct mmu_rb_handler *handler; atomic_t n_locked; - struct mm_struct *mm; }; struct hfi1_user_sdma_comp_q { @@ -250,4 +250,9 @@ int hfi1_user_sdma_process_request(struct hfi1_filedata *fd, struct iovec *iovec, unsigned long dim, unsigned long *count); +static inline struct mm_struct *mm_from_sdma_node(struct sdma_mmu_node *node) +{ + return node->rb.handler->mn.mm; +} + #endif /* _HFI1_USER_SDMA_H */

4 years

1
0
0 0

FAILED: patch "[PATCH] IB/hfi1: Ensure correct mm is used at all times" failed to apply to 4.14-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.14-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 3d2a9d642512c21a12d19b9250e7a835dcb41a79 Mon Sep 17 00:00:00 2001 From: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> Date: Wed, 25 Nov 2020 16:01:12 -0500 Subject: [PATCH] IB/hfi1: Ensure correct mm is used at all times Two earlier bug fixes have created a security problem in the hfi1 driver. One fix aimed to solve an issue where current->mm was not valid when closing the hfi1 cdev. It attempted to do this by saving a cached value of the current->mm pointer at file open time. This is a problem if another process with access to the FD calls in via write() or ioctl() to pin pages via the hfi driver. The other fix tried to solve a use after free by taking a reference on the mm. To fix this correctly we use the existing cached value of the mm in the mmu notifier. Now we can check in the insert, evict, etc. routines that current->mm matched what the notifier was registered for. If not, then don't allow access. The register of the mmu notifier will save the mm pointer. Since in do_exit() the exit_mm() is called before exit_files(), which would call our close routine a reference is needed on the mm. We rely on the mmgrab done by the registration of the notifier, whereas before it was explicit. The mmu notifier deregistration happens when the user context is torn down, the creation of which triggered the registration. Also of note is we do not do any explicit work to protect the interval tree notifier. It doesn't seem that this is going to be needed since we aren't actually doing anything with current->mm. The interval tree notifier stuff still has a FIXME noted from a previous commit that will be addressed in a follow on patch. Cc: <stable(a)vger.kernel.org> Fixes: e0cf75deab81 ("IB/hfi1: Fix mm_struct use after free") Fixes: 3faa3d9a308e ("IB/hfi1: Make use of mm consistent") Link: https://lore.kernel.org/r/20201125210112.104301.51331.stgit@awfm-01.aw.inte… Suggested-by: Jann Horn <jannh(a)google.com> Reported-by: Jason Gunthorpe <jgg(a)nvidia.com> Reviewed-by: Ira Weiny <ira.weiny(a)intel.com> Reviewed-by: Mike Marciniszyn <mike.marciniszyn(a)cornelisnetworks.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> Signed-off-by: Jason Gunthorpe <jgg(a)nvidia.com> diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c index 8ca51e43cf53..329ee4f48d95 100644 --- a/drivers/infiniband/hw/hfi1/file_ops.c +++ b/drivers/infiniband/hw/hfi1/file_ops.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2020 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -206,8 +207,6 @@ static int hfi1_file_open(struct inode *inode, struct file *fp) spin_lock_init(&fd->tid_lock); spin_lock_init(&fd->invalid_lock); fd->rec_cpu_num = -1; /* no cpu affinity by default */ - fd->mm = current->mm; - mmgrab(fd->mm); fd->dd = dd; fp->private_data = fd; return 0; @@ -711,7 +710,6 @@ static int hfi1_file_close(struct inode *inode, struct file *fp) deallocate_ctxt(uctxt); done: - mmdrop(fdata->mm); if (atomic_dec_and_test(&dd->user_refcount)) complete(&dd->user_comp); diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h index b4c6bff60a4e..e09e8244a94c 100644 --- a/drivers/infiniband/hw/hfi1/hfi.h +++ b/drivers/infiniband/hw/hfi1/hfi.h @@ -1,6 +1,7 @@ #ifndef _HFI1_KERNEL_H #define _HFI1_KERNEL_H /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2020 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -1451,7 +1452,6 @@ struct hfi1_filedata { u32 invalid_tid_idx; /* protect invalid_tids array and invalid_tid_idx */ spinlock_t invalid_lock; - struct mm_struct *mm; }; extern struct xarray hfi1_dev_table; diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.c b/drivers/infiniband/hw/hfi1/mmu_rb.c index 24ca17b77b72..f3fb28e3d5d7 100644 --- a/drivers/infiniband/hw/hfi1/mmu_rb.c +++ b/drivers/infiniband/hw/hfi1/mmu_rb.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2016 - 2017 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -48,23 +49,11 @@ #include <linux/rculist.h> #include <linux/mmu_notifier.h> #include <linux/interval_tree_generic.h> +#include <linux/sched/mm.h> #include "mmu_rb.h" #include "trace.h" -struct mmu_rb_handler { - struct mmu_notifier mn; - struct rb_root_cached root; - void *ops_arg; - spinlock_t lock; /* protect the RB tree */ - struct mmu_rb_ops *ops; - struct mm_struct *mm; - struct list_head lru_list; - struct work_struct del_work; - struct list_head del_list; - struct workqueue_struct *wq; -}; - static unsigned long mmu_node_start(struct mmu_rb_node *); static unsigned long mmu_node_last(struct mmu_rb_node *); static int mmu_notifier_range_start(struct mmu_notifier *, @@ -92,37 +81,36 @@ static unsigned long mmu_node_last(struct mmu_rb_node *node) return PAGE_ALIGN(node->addr + node->len) - 1; } -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm, +int hfi1_mmu_rb_register(void *ops_arg, struct mmu_rb_ops *ops, struct workqueue_struct *wq, struct mmu_rb_handler **handler) { - struct mmu_rb_handler *handlr; + struct mmu_rb_handler *h; int ret; - handlr = kmalloc(sizeof(*handlr), GFP_KERNEL); - if (!handlr) + h = kmalloc(sizeof(*h), GFP_KERNEL); + if (!h) return -ENOMEM; - handlr->root = RB_ROOT_CACHED; - handlr->ops = ops; - handlr->ops_arg = ops_arg; - INIT_HLIST_NODE(&handlr->mn.hlist); - spin_lock_init(&handlr->lock); - handlr->mn.ops = &mn_opts; - handlr->mm = mm; - INIT_WORK(&handlr->del_work, handle_remove); - INIT_LIST_HEAD(&handlr->del_list); - INIT_LIST_HEAD(&handlr->lru_list); - handlr->wq = wq; - - ret = mmu_notifier_register(&handlr->mn, handlr->mm); + h->root = RB_ROOT_CACHED; + h->ops = ops; + h->ops_arg = ops_arg; + INIT_HLIST_NODE(&h->mn.hlist); + spin_lock_init(&h->lock); + h->mn.ops = &mn_opts; + INIT_WORK(&h->del_work, handle_remove); + INIT_LIST_HEAD(&h->del_list); + INIT_LIST_HEAD(&h->lru_list); + h->wq = wq; + + ret = mmu_notifier_register(&h->mn, current->mm); if (ret) { - kfree(handlr); + kfree(h); return ret; } - *handler = handlr; + *handler = h; return 0; } @@ -134,7 +122,7 @@ void hfi1_mmu_rb_unregister(struct mmu_rb_handler *handler) struct list_head del_list; /* Unregister first so we don't get any more notifications. */ - mmu_notifier_unregister(&handler->mn, handler->mm); + mmu_notifier_unregister(&handler->mn, handler->mn.mm); /* * Make sure the wq delete handler is finished running. It will not @@ -166,6 +154,10 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler, int ret = 0; trace_hfi1_mmu_rb_insert(mnode->addr, mnode->len); + + if (current->mm != handler->mn.mm) + return -EPERM; + spin_lock_irqsave(&handler->lock, flags); node = __mmu_rb_search(handler, mnode->addr, mnode->len); if (node) { @@ -180,6 +172,7 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler, __mmu_int_rb_remove(mnode, &handler->root); list_del(&mnode->list); /* remove from LRU list */ } + mnode->handler = handler; unlock: spin_unlock_irqrestore(&handler->lock, flags); return ret; @@ -217,6 +210,9 @@ bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler, unsigned long flags; bool ret = false; + if (current->mm != handler->mn.mm) + return ret; + spin_lock_irqsave(&handler->lock, flags); node = __mmu_rb_search(handler, addr, len); if (node) { @@ -239,6 +235,9 @@ void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg) unsigned long flags; bool stop = false; + if (current->mm != handler->mn.mm) + return; + INIT_LIST_HEAD(&del_list); spin_lock_irqsave(&handler->lock, flags); @@ -272,6 +271,9 @@ void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler, { unsigned long flags; + if (current->mm != handler->mn.mm) + return; + /* Validity of handler and node pointers has been checked by caller. */ trace_hfi1_mmu_rb_remove(node->addr, node->len); spin_lock_irqsave(&handler->lock, flags); diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.h b/drivers/infiniband/hw/hfi1/mmu_rb.h index f04cec1e99d1..423aacc67e94 100644 --- a/drivers/infiniband/hw/hfi1/mmu_rb.h +++ b/drivers/infiniband/hw/hfi1/mmu_rb.h @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2016 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -54,6 +55,7 @@ struct mmu_rb_node { unsigned long len; unsigned long __last; struct rb_node node; + struct mmu_rb_handler *handler; struct list_head list; }; @@ -71,7 +73,19 @@ struct mmu_rb_ops { void *evict_arg, bool *stop); }; -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm, +struct mmu_rb_handler { + struct mmu_notifier mn; + struct rb_root_cached root; + void *ops_arg; + spinlock_t lock; /* protect the RB tree */ + struct mmu_rb_ops *ops; + struct list_head lru_list; + struct work_struct del_work; + struct list_head del_list; + struct workqueue_struct *wq; +}; + +int hfi1_mmu_rb_register(void *ops_arg, struct mmu_rb_ops *ops, struct workqueue_struct *wq, struct mmu_rb_handler **handler); diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.c b/drivers/infiniband/hw/hfi1/user_exp_rcv.c index f81ca20f4b69..b94fc7fd75a9 100644 --- a/drivers/infiniband/hw/hfi1/user_exp_rcv.c +++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -173,15 +174,18 @@ static void unpin_rcv_pages(struct hfi1_filedata *fd, { struct page **pages; struct hfi1_devdata *dd = fd->uctxt->dd; + struct mm_struct *mm; if (mapped) { pci_unmap_single(dd->pcidev, node->dma_addr, node->npages * PAGE_SIZE, PCI_DMA_FROMDEVICE); pages = &node->pages[idx]; + mm = mm_from_tid_node(node); } else { pages = &tidbuf->pages[idx]; + mm = current->mm; } - hfi1_release_user_pages(fd->mm, pages, npages, mapped); + hfi1_release_user_pages(mm, pages, npages, mapped); fd->tid_n_pinned -= npages; } @@ -216,12 +220,12 @@ static int pin_rcv_pages(struct hfi1_filedata *fd, struct tid_user_buf *tidbuf) * pages, accept the amount pinned so far and program only that. * User space knows how to deal with partially programmed buffers. */ - if (!hfi1_can_pin_pages(dd, fd->mm, fd->tid_n_pinned, npages)) { + if (!hfi1_can_pin_pages(dd, current->mm, fd->tid_n_pinned, npages)) { kfree(pages); return -ENOMEM; } - pinned = hfi1_acquire_user_pages(fd->mm, vaddr, npages, true, pages); + pinned = hfi1_acquire_user_pages(current->mm, vaddr, npages, true, pages); if (pinned <= 0) { kfree(pages); return pinned; @@ -756,7 +760,7 @@ static int set_rcvarray_entry(struct hfi1_filedata *fd, if (fd->use_mn) { ret = mmu_interval_notifier_insert( - &node->notifier, fd->mm, + &node->notifier, current->mm, tbuf->vaddr + (pageidx * PAGE_SIZE), npages * PAGE_SIZE, &tid_mn_ops); if (ret) diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.h b/drivers/infiniband/hw/hfi1/user_exp_rcv.h index 332abb446861..d45c7b6988d4 100644 --- a/drivers/infiniband/hw/hfi1/user_exp_rcv.h +++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.h @@ -1,6 +1,7 @@ #ifndef _HFI1_USER_EXP_RCV_H #define _HFI1_USER_EXP_RCV_H /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2017 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -95,4 +96,9 @@ int hfi1_user_exp_rcv_clear(struct hfi1_filedata *fd, int hfi1_user_exp_rcv_invalid(struct hfi1_filedata *fd, struct hfi1_tid_info *tinfo); +static inline struct mm_struct *mm_from_tid_node(struct tid_rb_node *node) +{ + return node->notifier.mm; +} + #endif /* _HFI1_USER_EXP_RCV_H */ diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c index a92346e88628..4a4956f96a7e 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.c +++ b/drivers/infiniband/hw/hfi1/user_sdma.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -188,7 +189,6 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, atomic_set(&pq->n_reqs, 0); init_waitqueue_head(&pq->wait); atomic_set(&pq->n_locked, 0); - pq->mm = fd->mm; iowait_init(&pq->busy, 0, NULL, NULL, defer_packet_queue, activate_packet_queue, NULL, NULL); @@ -230,7 +230,7 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, cq->nentries = hfi1_sdma_comp_ring_size; - ret = hfi1_mmu_rb_register(pq, pq->mm, &sdma_rb_ops, dd->pport->hfi1_wq, + ret = hfi1_mmu_rb_register(pq, &sdma_rb_ops, dd->pport->hfi1_wq, &pq->handler); if (ret) { dd_dev_err(dd, "Failed to register with MMU %d", ret); @@ -980,13 +980,13 @@ static int pin_sdma_pages(struct user_sdma_request *req, npages -= node->npages; retry: - if (!hfi1_can_pin_pages(pq->dd, pq->mm, + if (!hfi1_can_pin_pages(pq->dd, current->mm, atomic_read(&pq->n_locked), npages)) { cleared = sdma_cache_evict(pq, npages); if (cleared >= npages) goto retry; } - pinned = hfi1_acquire_user_pages(pq->mm, + pinned = hfi1_acquire_user_pages(current->mm, ((unsigned long)iovec->iov.iov_base + (node->npages * PAGE_SIZE)), npages, 0, pages + node->npages); @@ -995,7 +995,7 @@ static int pin_sdma_pages(struct user_sdma_request *req, return pinned; } if (pinned != npages) { - unpin_vector_pages(pq->mm, pages, node->npages, pinned); + unpin_vector_pages(current->mm, pages, node->npages, pinned); return -EFAULT; } kfree(node->pages); @@ -1008,7 +1008,8 @@ static int pin_sdma_pages(struct user_sdma_request *req, static void unpin_sdma_pages(struct sdma_mmu_node *node) { if (node->npages) { - unpin_vector_pages(node->pq->mm, node->pages, 0, node->npages); + unpin_vector_pages(mm_from_sdma_node(node), node->pages, 0, + node->npages); atomic_sub(node->npages, &node->pq->n_locked); } } diff --git a/drivers/infiniband/hw/hfi1/user_sdma.h b/drivers/infiniband/hw/hfi1/user_sdma.h index 9972e0e6545e..1e8c02fe8ad1 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.h +++ b/drivers/infiniband/hw/hfi1/user_sdma.h @@ -1,6 +1,7 @@ #ifndef _HFI1_USER_SDMA_H #define _HFI1_USER_SDMA_H /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -133,7 +134,6 @@ struct hfi1_user_sdma_pkt_q { unsigned long unpinned; struct mmu_rb_handler *handler; atomic_t n_locked; - struct mm_struct *mm; }; struct hfi1_user_sdma_comp_q { @@ -250,4 +250,9 @@ int hfi1_user_sdma_process_request(struct hfi1_filedata *fd, struct iovec *iovec, unsigned long dim, unsigned long *count); +static inline struct mm_struct *mm_from_sdma_node(struct sdma_mmu_node *node) +{ + return node->rb.handler->mn.mm; +} + #endif /* _HFI1_USER_SDMA_H */

4 years

1
0
0 0

FAILED: patch "[PATCH] IB/hfi1: Ensure correct mm is used at all times" failed to apply to 4.19-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 4.19-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 3d2a9d642512c21a12d19b9250e7a835dcb41a79 Mon Sep 17 00:00:00 2001 From: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> Date: Wed, 25 Nov 2020 16:01:12 -0500 Subject: [PATCH] IB/hfi1: Ensure correct mm is used at all times Two earlier bug fixes have created a security problem in the hfi1 driver. One fix aimed to solve an issue where current->mm was not valid when closing the hfi1 cdev. It attempted to do this by saving a cached value of the current->mm pointer at file open time. This is a problem if another process with access to the FD calls in via write() or ioctl() to pin pages via the hfi driver. The other fix tried to solve a use after free by taking a reference on the mm. To fix this correctly we use the existing cached value of the mm in the mmu notifier. Now we can check in the insert, evict, etc. routines that current->mm matched what the notifier was registered for. If not, then don't allow access. The register of the mmu notifier will save the mm pointer. Since in do_exit() the exit_mm() is called before exit_files(), which would call our close routine a reference is needed on the mm. We rely on the mmgrab done by the registration of the notifier, whereas before it was explicit. The mmu notifier deregistration happens when the user context is torn down, the creation of which triggered the registration. Also of note is we do not do any explicit work to protect the interval tree notifier. It doesn't seem that this is going to be needed since we aren't actually doing anything with current->mm. The interval tree notifier stuff still has a FIXME noted from a previous commit that will be addressed in a follow on patch. Cc: <stable(a)vger.kernel.org> Fixes: e0cf75deab81 ("IB/hfi1: Fix mm_struct use after free") Fixes: 3faa3d9a308e ("IB/hfi1: Make use of mm consistent") Link: https://lore.kernel.org/r/20201125210112.104301.51331.stgit@awfm-01.aw.inte… Suggested-by: Jann Horn <jannh(a)google.com> Reported-by: Jason Gunthorpe <jgg(a)nvidia.com> Reviewed-by: Ira Weiny <ira.weiny(a)intel.com> Reviewed-by: Mike Marciniszyn <mike.marciniszyn(a)cornelisnetworks.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> Signed-off-by: Jason Gunthorpe <jgg(a)nvidia.com> diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c index 8ca51e43cf53..329ee4f48d95 100644 --- a/drivers/infiniband/hw/hfi1/file_ops.c +++ b/drivers/infiniband/hw/hfi1/file_ops.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2020 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -206,8 +207,6 @@ static int hfi1_file_open(struct inode *inode, struct file *fp) spin_lock_init(&fd->tid_lock); spin_lock_init(&fd->invalid_lock); fd->rec_cpu_num = -1; /* no cpu affinity by default */ - fd->mm = current->mm; - mmgrab(fd->mm); fd->dd = dd; fp->private_data = fd; return 0; @@ -711,7 +710,6 @@ static int hfi1_file_close(struct inode *inode, struct file *fp) deallocate_ctxt(uctxt); done: - mmdrop(fdata->mm); if (atomic_dec_and_test(&dd->user_refcount)) complete(&dd->user_comp); diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h index b4c6bff60a4e..e09e8244a94c 100644 --- a/drivers/infiniband/hw/hfi1/hfi.h +++ b/drivers/infiniband/hw/hfi1/hfi.h @@ -1,6 +1,7 @@ #ifndef _HFI1_KERNEL_H #define _HFI1_KERNEL_H /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2020 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -1451,7 +1452,6 @@ struct hfi1_filedata { u32 invalid_tid_idx; /* protect invalid_tids array and invalid_tid_idx */ spinlock_t invalid_lock; - struct mm_struct *mm; }; extern struct xarray hfi1_dev_table; diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.c b/drivers/infiniband/hw/hfi1/mmu_rb.c index 24ca17b77b72..f3fb28e3d5d7 100644 --- a/drivers/infiniband/hw/hfi1/mmu_rb.c +++ b/drivers/infiniband/hw/hfi1/mmu_rb.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2016 - 2017 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -48,23 +49,11 @@ #include <linux/rculist.h> #include <linux/mmu_notifier.h> #include <linux/interval_tree_generic.h> +#include <linux/sched/mm.h> #include "mmu_rb.h" #include "trace.h" -struct mmu_rb_handler { - struct mmu_notifier mn; - struct rb_root_cached root; - void *ops_arg; - spinlock_t lock; /* protect the RB tree */ - struct mmu_rb_ops *ops; - struct mm_struct *mm; - struct list_head lru_list; - struct work_struct del_work; - struct list_head del_list; - struct workqueue_struct *wq; -}; - static unsigned long mmu_node_start(struct mmu_rb_node *); static unsigned long mmu_node_last(struct mmu_rb_node *); static int mmu_notifier_range_start(struct mmu_notifier *, @@ -92,37 +81,36 @@ static unsigned long mmu_node_last(struct mmu_rb_node *node) return PAGE_ALIGN(node->addr + node->len) - 1; } -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm, +int hfi1_mmu_rb_register(void *ops_arg, struct mmu_rb_ops *ops, struct workqueue_struct *wq, struct mmu_rb_handler **handler) { - struct mmu_rb_handler *handlr; + struct mmu_rb_handler *h; int ret; - handlr = kmalloc(sizeof(*handlr), GFP_KERNEL); - if (!handlr) + h = kmalloc(sizeof(*h), GFP_KERNEL); + if (!h) return -ENOMEM; - handlr->root = RB_ROOT_CACHED; - handlr->ops = ops; - handlr->ops_arg = ops_arg; - INIT_HLIST_NODE(&handlr->mn.hlist); - spin_lock_init(&handlr->lock); - handlr->mn.ops = &mn_opts; - handlr->mm = mm; - INIT_WORK(&handlr->del_work, handle_remove); - INIT_LIST_HEAD(&handlr->del_list); - INIT_LIST_HEAD(&handlr->lru_list); - handlr->wq = wq; - - ret = mmu_notifier_register(&handlr->mn, handlr->mm); + h->root = RB_ROOT_CACHED; + h->ops = ops; + h->ops_arg = ops_arg; + INIT_HLIST_NODE(&h->mn.hlist); + spin_lock_init(&h->lock); + h->mn.ops = &mn_opts; + INIT_WORK(&h->del_work, handle_remove); + INIT_LIST_HEAD(&h->del_list); + INIT_LIST_HEAD(&h->lru_list); + h->wq = wq; + + ret = mmu_notifier_register(&h->mn, current->mm); if (ret) { - kfree(handlr); + kfree(h); return ret; } - *handler = handlr; + *handler = h; return 0; } @@ -134,7 +122,7 @@ void hfi1_mmu_rb_unregister(struct mmu_rb_handler *handler) struct list_head del_list; /* Unregister first so we don't get any more notifications. */ - mmu_notifier_unregister(&handler->mn, handler->mm); + mmu_notifier_unregister(&handler->mn, handler->mn.mm); /* * Make sure the wq delete handler is finished running. It will not @@ -166,6 +154,10 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler, int ret = 0; trace_hfi1_mmu_rb_insert(mnode->addr, mnode->len); + + if (current->mm != handler->mn.mm) + return -EPERM; + spin_lock_irqsave(&handler->lock, flags); node = __mmu_rb_search(handler, mnode->addr, mnode->len); if (node) { @@ -180,6 +172,7 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler, __mmu_int_rb_remove(mnode, &handler->root); list_del(&mnode->list); /* remove from LRU list */ } + mnode->handler = handler; unlock: spin_unlock_irqrestore(&handler->lock, flags); return ret; @@ -217,6 +210,9 @@ bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler, unsigned long flags; bool ret = false; + if (current->mm != handler->mn.mm) + return ret; + spin_lock_irqsave(&handler->lock, flags); node = __mmu_rb_search(handler, addr, len); if (node) { @@ -239,6 +235,9 @@ void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg) unsigned long flags; bool stop = false; + if (current->mm != handler->mn.mm) + return; + INIT_LIST_HEAD(&del_list); spin_lock_irqsave(&handler->lock, flags); @@ -272,6 +271,9 @@ void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler, { unsigned long flags; + if (current->mm != handler->mn.mm) + return; + /* Validity of handler and node pointers has been checked by caller. */ trace_hfi1_mmu_rb_remove(node->addr, node->len); spin_lock_irqsave(&handler->lock, flags); diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.h b/drivers/infiniband/hw/hfi1/mmu_rb.h index f04cec1e99d1..423aacc67e94 100644 --- a/drivers/infiniband/hw/hfi1/mmu_rb.h +++ b/drivers/infiniband/hw/hfi1/mmu_rb.h @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2016 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -54,6 +55,7 @@ struct mmu_rb_node { unsigned long len; unsigned long __last; struct rb_node node; + struct mmu_rb_handler *handler; struct list_head list; }; @@ -71,7 +73,19 @@ struct mmu_rb_ops { void *evict_arg, bool *stop); }; -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm, +struct mmu_rb_handler { + struct mmu_notifier mn; + struct rb_root_cached root; + void *ops_arg; + spinlock_t lock; /* protect the RB tree */ + struct mmu_rb_ops *ops; + struct list_head lru_list; + struct work_struct del_work; + struct list_head del_list; + struct workqueue_struct *wq; +}; + +int hfi1_mmu_rb_register(void *ops_arg, struct mmu_rb_ops *ops, struct workqueue_struct *wq, struct mmu_rb_handler **handler); diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.c b/drivers/infiniband/hw/hfi1/user_exp_rcv.c index f81ca20f4b69..b94fc7fd75a9 100644 --- a/drivers/infiniband/hw/hfi1/user_exp_rcv.c +++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -173,15 +174,18 @@ static void unpin_rcv_pages(struct hfi1_filedata *fd, { struct page **pages; struct hfi1_devdata *dd = fd->uctxt->dd; + struct mm_struct *mm; if (mapped) { pci_unmap_single(dd->pcidev, node->dma_addr, node->npages * PAGE_SIZE, PCI_DMA_FROMDEVICE); pages = &node->pages[idx]; + mm = mm_from_tid_node(node); } else { pages = &tidbuf->pages[idx]; + mm = current->mm; } - hfi1_release_user_pages(fd->mm, pages, npages, mapped); + hfi1_release_user_pages(mm, pages, npages, mapped); fd->tid_n_pinned -= npages; } @@ -216,12 +220,12 @@ static int pin_rcv_pages(struct hfi1_filedata *fd, struct tid_user_buf *tidbuf) * pages, accept the amount pinned so far and program only that. * User space knows how to deal with partially programmed buffers. */ - if (!hfi1_can_pin_pages(dd, fd->mm, fd->tid_n_pinned, npages)) { + if (!hfi1_can_pin_pages(dd, current->mm, fd->tid_n_pinned, npages)) { kfree(pages); return -ENOMEM; } - pinned = hfi1_acquire_user_pages(fd->mm, vaddr, npages, true, pages); + pinned = hfi1_acquire_user_pages(current->mm, vaddr, npages, true, pages); if (pinned <= 0) { kfree(pages); return pinned; @@ -756,7 +760,7 @@ static int set_rcvarray_entry(struct hfi1_filedata *fd, if (fd->use_mn) { ret = mmu_interval_notifier_insert( - &node->notifier, fd->mm, + &node->notifier, current->mm, tbuf->vaddr + (pageidx * PAGE_SIZE), npages * PAGE_SIZE, &tid_mn_ops); if (ret) diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.h b/drivers/infiniband/hw/hfi1/user_exp_rcv.h index 332abb446861..d45c7b6988d4 100644 --- a/drivers/infiniband/hw/hfi1/user_exp_rcv.h +++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.h @@ -1,6 +1,7 @@ #ifndef _HFI1_USER_EXP_RCV_H #define _HFI1_USER_EXP_RCV_H /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2017 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -95,4 +96,9 @@ int hfi1_user_exp_rcv_clear(struct hfi1_filedata *fd, int hfi1_user_exp_rcv_invalid(struct hfi1_filedata *fd, struct hfi1_tid_info *tinfo); +static inline struct mm_struct *mm_from_tid_node(struct tid_rb_node *node) +{ + return node->notifier.mm; +} + #endif /* _HFI1_USER_EXP_RCV_H */ diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c index a92346e88628..4a4956f96a7e 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.c +++ b/drivers/infiniband/hw/hfi1/user_sdma.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -188,7 +189,6 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, atomic_set(&pq->n_reqs, 0); init_waitqueue_head(&pq->wait); atomic_set(&pq->n_locked, 0); - pq->mm = fd->mm; iowait_init(&pq->busy, 0, NULL, NULL, defer_packet_queue, activate_packet_queue, NULL, NULL); @@ -230,7 +230,7 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, cq->nentries = hfi1_sdma_comp_ring_size; - ret = hfi1_mmu_rb_register(pq, pq->mm, &sdma_rb_ops, dd->pport->hfi1_wq, + ret = hfi1_mmu_rb_register(pq, &sdma_rb_ops, dd->pport->hfi1_wq, &pq->handler); if (ret) { dd_dev_err(dd, "Failed to register with MMU %d", ret); @@ -980,13 +980,13 @@ static int pin_sdma_pages(struct user_sdma_request *req, npages -= node->npages; retry: - if (!hfi1_can_pin_pages(pq->dd, pq->mm, + if (!hfi1_can_pin_pages(pq->dd, current->mm, atomic_read(&pq->n_locked), npages)) { cleared = sdma_cache_evict(pq, npages); if (cleared >= npages) goto retry; } - pinned = hfi1_acquire_user_pages(pq->mm, + pinned = hfi1_acquire_user_pages(current->mm, ((unsigned long)iovec->iov.iov_base + (node->npages * PAGE_SIZE)), npages, 0, pages + node->npages); @@ -995,7 +995,7 @@ static int pin_sdma_pages(struct user_sdma_request *req, return pinned; } if (pinned != npages) { - unpin_vector_pages(pq->mm, pages, node->npages, pinned); + unpin_vector_pages(current->mm, pages, node->npages, pinned); return -EFAULT; } kfree(node->pages); @@ -1008,7 +1008,8 @@ static int pin_sdma_pages(struct user_sdma_request *req, static void unpin_sdma_pages(struct sdma_mmu_node *node) { if (node->npages) { - unpin_vector_pages(node->pq->mm, node->pages, 0, node->npages); + unpin_vector_pages(mm_from_sdma_node(node), node->pages, 0, + node->npages); atomic_sub(node->npages, &node->pq->n_locked); } } diff --git a/drivers/infiniband/hw/hfi1/user_sdma.h b/drivers/infiniband/hw/hfi1/user_sdma.h index 9972e0e6545e..1e8c02fe8ad1 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.h +++ b/drivers/infiniband/hw/hfi1/user_sdma.h @@ -1,6 +1,7 @@ #ifndef _HFI1_USER_SDMA_H #define _HFI1_USER_SDMA_H /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -133,7 +134,6 @@ struct hfi1_user_sdma_pkt_q { unsigned long unpinned; struct mmu_rb_handler *handler; atomic_t n_locked; - struct mm_struct *mm; }; struct hfi1_user_sdma_comp_q { @@ -250,4 +250,9 @@ int hfi1_user_sdma_process_request(struct hfi1_filedata *fd, struct iovec *iovec, unsigned long dim, unsigned long *count); +static inline struct mm_struct *mm_from_sdma_node(struct sdma_mmu_node *node) +{ + return node->rb.handler->mn.mm; +} + #endif /* _HFI1_USER_SDMA_H */

4 years

1
0
0 0

FAILED: patch "[PATCH] IB/hfi1: Ensure correct mm is used at all times" failed to apply to 5.4-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 5.4-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 3d2a9d642512c21a12d19b9250e7a835dcb41a79 Mon Sep 17 00:00:00 2001 From: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> Date: Wed, 25 Nov 2020 16:01:12 -0500 Subject: [PATCH] IB/hfi1: Ensure correct mm is used at all times Two earlier bug fixes have created a security problem in the hfi1 driver. One fix aimed to solve an issue where current->mm was not valid when closing the hfi1 cdev. It attempted to do this by saving a cached value of the current->mm pointer at file open time. This is a problem if another process with access to the FD calls in via write() or ioctl() to pin pages via the hfi driver. The other fix tried to solve a use after free by taking a reference on the mm. To fix this correctly we use the existing cached value of the mm in the mmu notifier. Now we can check in the insert, evict, etc. routines that current->mm matched what the notifier was registered for. If not, then don't allow access. The register of the mmu notifier will save the mm pointer. Since in do_exit() the exit_mm() is called before exit_files(), which would call our close routine a reference is needed on the mm. We rely on the mmgrab done by the registration of the notifier, whereas before it was explicit. The mmu notifier deregistration happens when the user context is torn down, the creation of which triggered the registration. Also of note is we do not do any explicit work to protect the interval tree notifier. It doesn't seem that this is going to be needed since we aren't actually doing anything with current->mm. The interval tree notifier stuff still has a FIXME noted from a previous commit that will be addressed in a follow on patch. Cc: <stable(a)vger.kernel.org> Fixes: e0cf75deab81 ("IB/hfi1: Fix mm_struct use after free") Fixes: 3faa3d9a308e ("IB/hfi1: Make use of mm consistent") Link: https://lore.kernel.org/r/20201125210112.104301.51331.stgit@awfm-01.aw.inte… Suggested-by: Jann Horn <jannh(a)google.com> Reported-by: Jason Gunthorpe <jgg(a)nvidia.com> Reviewed-by: Ira Weiny <ira.weiny(a)intel.com> Reviewed-by: Mike Marciniszyn <mike.marciniszyn(a)cornelisnetworks.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> Signed-off-by: Jason Gunthorpe <jgg(a)nvidia.com> diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c index 8ca51e43cf53..329ee4f48d95 100644 --- a/drivers/infiniband/hw/hfi1/file_ops.c +++ b/drivers/infiniband/hw/hfi1/file_ops.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2020 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -206,8 +207,6 @@ static int hfi1_file_open(struct inode *inode, struct file *fp) spin_lock_init(&fd->tid_lock); spin_lock_init(&fd->invalid_lock); fd->rec_cpu_num = -1; /* no cpu affinity by default */ - fd->mm = current->mm; - mmgrab(fd->mm); fd->dd = dd; fp->private_data = fd; return 0; @@ -711,7 +710,6 @@ static int hfi1_file_close(struct inode *inode, struct file *fp) deallocate_ctxt(uctxt); done: - mmdrop(fdata->mm); if (atomic_dec_and_test(&dd->user_refcount)) complete(&dd->user_comp); diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h index b4c6bff60a4e..e09e8244a94c 100644 --- a/drivers/infiniband/hw/hfi1/hfi.h +++ b/drivers/infiniband/hw/hfi1/hfi.h @@ -1,6 +1,7 @@ #ifndef _HFI1_KERNEL_H #define _HFI1_KERNEL_H /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2020 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -1451,7 +1452,6 @@ struct hfi1_filedata { u32 invalid_tid_idx; /* protect invalid_tids array and invalid_tid_idx */ spinlock_t invalid_lock; - struct mm_struct *mm; }; extern struct xarray hfi1_dev_table; diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.c b/drivers/infiniband/hw/hfi1/mmu_rb.c index 24ca17b77b72..f3fb28e3d5d7 100644 --- a/drivers/infiniband/hw/hfi1/mmu_rb.c +++ b/drivers/infiniband/hw/hfi1/mmu_rb.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2016 - 2017 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -48,23 +49,11 @@ #include <linux/rculist.h> #include <linux/mmu_notifier.h> #include <linux/interval_tree_generic.h> +#include <linux/sched/mm.h> #include "mmu_rb.h" #include "trace.h" -struct mmu_rb_handler { - struct mmu_notifier mn; - struct rb_root_cached root; - void *ops_arg; - spinlock_t lock; /* protect the RB tree */ - struct mmu_rb_ops *ops; - struct mm_struct *mm; - struct list_head lru_list; - struct work_struct del_work; - struct list_head del_list; - struct workqueue_struct *wq; -}; - static unsigned long mmu_node_start(struct mmu_rb_node *); static unsigned long mmu_node_last(struct mmu_rb_node *); static int mmu_notifier_range_start(struct mmu_notifier *, @@ -92,37 +81,36 @@ static unsigned long mmu_node_last(struct mmu_rb_node *node) return PAGE_ALIGN(node->addr + node->len) - 1; } -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm, +int hfi1_mmu_rb_register(void *ops_arg, struct mmu_rb_ops *ops, struct workqueue_struct *wq, struct mmu_rb_handler **handler) { - struct mmu_rb_handler *handlr; + struct mmu_rb_handler *h; int ret; - handlr = kmalloc(sizeof(*handlr), GFP_KERNEL); - if (!handlr) + h = kmalloc(sizeof(*h), GFP_KERNEL); + if (!h) return -ENOMEM; - handlr->root = RB_ROOT_CACHED; - handlr->ops = ops; - handlr->ops_arg = ops_arg; - INIT_HLIST_NODE(&handlr->mn.hlist); - spin_lock_init(&handlr->lock); - handlr->mn.ops = &mn_opts; - handlr->mm = mm; - INIT_WORK(&handlr->del_work, handle_remove); - INIT_LIST_HEAD(&handlr->del_list); - INIT_LIST_HEAD(&handlr->lru_list); - handlr->wq = wq; - - ret = mmu_notifier_register(&handlr->mn, handlr->mm); + h->root = RB_ROOT_CACHED; + h->ops = ops; + h->ops_arg = ops_arg; + INIT_HLIST_NODE(&h->mn.hlist); + spin_lock_init(&h->lock); + h->mn.ops = &mn_opts; + INIT_WORK(&h->del_work, handle_remove); + INIT_LIST_HEAD(&h->del_list); + INIT_LIST_HEAD(&h->lru_list); + h->wq = wq; + + ret = mmu_notifier_register(&h->mn, current->mm); if (ret) { - kfree(handlr); + kfree(h); return ret; } - *handler = handlr; + *handler = h; return 0; } @@ -134,7 +122,7 @@ void hfi1_mmu_rb_unregister(struct mmu_rb_handler *handler) struct list_head del_list; /* Unregister first so we don't get any more notifications. */ - mmu_notifier_unregister(&handler->mn, handler->mm); + mmu_notifier_unregister(&handler->mn, handler->mn.mm); /* * Make sure the wq delete handler is finished running. It will not @@ -166,6 +154,10 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler, int ret = 0; trace_hfi1_mmu_rb_insert(mnode->addr, mnode->len); + + if (current->mm != handler->mn.mm) + return -EPERM; + spin_lock_irqsave(&handler->lock, flags); node = __mmu_rb_search(handler, mnode->addr, mnode->len); if (node) { @@ -180,6 +172,7 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler, __mmu_int_rb_remove(mnode, &handler->root); list_del(&mnode->list); /* remove from LRU list */ } + mnode->handler = handler; unlock: spin_unlock_irqrestore(&handler->lock, flags); return ret; @@ -217,6 +210,9 @@ bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler, unsigned long flags; bool ret = false; + if (current->mm != handler->mn.mm) + return ret; + spin_lock_irqsave(&handler->lock, flags); node = __mmu_rb_search(handler, addr, len); if (node) { @@ -239,6 +235,9 @@ void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg) unsigned long flags; bool stop = false; + if (current->mm != handler->mn.mm) + return; + INIT_LIST_HEAD(&del_list); spin_lock_irqsave(&handler->lock, flags); @@ -272,6 +271,9 @@ void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler, { unsigned long flags; + if (current->mm != handler->mn.mm) + return; + /* Validity of handler and node pointers has been checked by caller. */ trace_hfi1_mmu_rb_remove(node->addr, node->len); spin_lock_irqsave(&handler->lock, flags); diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.h b/drivers/infiniband/hw/hfi1/mmu_rb.h index f04cec1e99d1..423aacc67e94 100644 --- a/drivers/infiniband/hw/hfi1/mmu_rb.h +++ b/drivers/infiniband/hw/hfi1/mmu_rb.h @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2016 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -54,6 +55,7 @@ struct mmu_rb_node { unsigned long len; unsigned long __last; struct rb_node node; + struct mmu_rb_handler *handler; struct list_head list; }; @@ -71,7 +73,19 @@ struct mmu_rb_ops { void *evict_arg, bool *stop); }; -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm, +struct mmu_rb_handler { + struct mmu_notifier mn; + struct rb_root_cached root; + void *ops_arg; + spinlock_t lock; /* protect the RB tree */ + struct mmu_rb_ops *ops; + struct list_head lru_list; + struct work_struct del_work; + struct list_head del_list; + struct workqueue_struct *wq; +}; + +int hfi1_mmu_rb_register(void *ops_arg, struct mmu_rb_ops *ops, struct workqueue_struct *wq, struct mmu_rb_handler **handler); diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.c b/drivers/infiniband/hw/hfi1/user_exp_rcv.c index f81ca20f4b69..b94fc7fd75a9 100644 --- a/drivers/infiniband/hw/hfi1/user_exp_rcv.c +++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 Cornelis Networks, Inc. * Copyright(c) 2015-2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -173,15 +174,18 @@ static void unpin_rcv_pages(struct hfi1_filedata *fd, { struct page **pages; struct hfi1_devdata *dd = fd->uctxt->dd; + struct mm_struct *mm; if (mapped) { pci_unmap_single(dd->pcidev, node->dma_addr, node->npages * PAGE_SIZE, PCI_DMA_FROMDEVICE); pages = &node->pages[idx]; + mm = mm_from_tid_node(node); } else { pages = &tidbuf->pages[idx]; + mm = current->mm; } - hfi1_release_user_pages(fd->mm, pages, npages, mapped); + hfi1_release_user_pages(mm, pages, npages, mapped); fd->tid_n_pinned -= npages; } @@ -216,12 +220,12 @@ static int pin_rcv_pages(struct hfi1_filedata *fd, struct tid_user_buf *tidbuf) * pages, accept the amount pinned so far and program only that. * User space knows how to deal with partially programmed buffers. */ - if (!hfi1_can_pin_pages(dd, fd->mm, fd->tid_n_pinned, npages)) { + if (!hfi1_can_pin_pages(dd, current->mm, fd->tid_n_pinned, npages)) { kfree(pages); return -ENOMEM; } - pinned = hfi1_acquire_user_pages(fd->mm, vaddr, npages, true, pages); + pinned = hfi1_acquire_user_pages(current->mm, vaddr, npages, true, pages); if (pinned <= 0) { kfree(pages); return pinned; @@ -756,7 +760,7 @@ static int set_rcvarray_entry(struct hfi1_filedata *fd, if (fd->use_mn) { ret = mmu_interval_notifier_insert( - &node->notifier, fd->mm, + &node->notifier, current->mm, tbuf->vaddr + (pageidx * PAGE_SIZE), npages * PAGE_SIZE, &tid_mn_ops); if (ret) diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.h b/drivers/infiniband/hw/hfi1/user_exp_rcv.h index 332abb446861..d45c7b6988d4 100644 --- a/drivers/infiniband/hw/hfi1/user_exp_rcv.h +++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.h @@ -1,6 +1,7 @@ #ifndef _HFI1_USER_EXP_RCV_H #define _HFI1_USER_EXP_RCV_H /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2017 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -95,4 +96,9 @@ int hfi1_user_exp_rcv_clear(struct hfi1_filedata *fd, int hfi1_user_exp_rcv_invalid(struct hfi1_filedata *fd, struct hfi1_tid_info *tinfo); +static inline struct mm_struct *mm_from_tid_node(struct tid_rb_node *node) +{ + return node->notifier.mm; +} + #endif /* _HFI1_USER_EXP_RCV_H */ diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c index a92346e88628..4a4956f96a7e 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.c +++ b/drivers/infiniband/hw/hfi1/user_sdma.c @@ -1,4 +1,5 @@ /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -188,7 +189,6 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, atomic_set(&pq->n_reqs, 0); init_waitqueue_head(&pq->wait); atomic_set(&pq->n_locked, 0); - pq->mm = fd->mm; iowait_init(&pq->busy, 0, NULL, NULL, defer_packet_queue, activate_packet_queue, NULL, NULL); @@ -230,7 +230,7 @@ int hfi1_user_sdma_alloc_queues(struct hfi1_ctxtdata *uctxt, cq->nentries = hfi1_sdma_comp_ring_size; - ret = hfi1_mmu_rb_register(pq, pq->mm, &sdma_rb_ops, dd->pport->hfi1_wq, + ret = hfi1_mmu_rb_register(pq, &sdma_rb_ops, dd->pport->hfi1_wq, &pq->handler); if (ret) { dd_dev_err(dd, "Failed to register with MMU %d", ret); @@ -980,13 +980,13 @@ static int pin_sdma_pages(struct user_sdma_request *req, npages -= node->npages; retry: - if (!hfi1_can_pin_pages(pq->dd, pq->mm, + if (!hfi1_can_pin_pages(pq->dd, current->mm, atomic_read(&pq->n_locked), npages)) { cleared = sdma_cache_evict(pq, npages); if (cleared >= npages) goto retry; } - pinned = hfi1_acquire_user_pages(pq->mm, + pinned = hfi1_acquire_user_pages(current->mm, ((unsigned long)iovec->iov.iov_base + (node->npages * PAGE_SIZE)), npages, 0, pages + node->npages); @@ -995,7 +995,7 @@ static int pin_sdma_pages(struct user_sdma_request *req, return pinned; } if (pinned != npages) { - unpin_vector_pages(pq->mm, pages, node->npages, pinned); + unpin_vector_pages(current->mm, pages, node->npages, pinned); return -EFAULT; } kfree(node->pages); @@ -1008,7 +1008,8 @@ static int pin_sdma_pages(struct user_sdma_request *req, static void unpin_sdma_pages(struct sdma_mmu_node *node) { if (node->npages) { - unpin_vector_pages(node->pq->mm, node->pages, 0, node->npages); + unpin_vector_pages(mm_from_sdma_node(node), node->pages, 0, + node->npages); atomic_sub(node->npages, &node->pq->n_locked); } } diff --git a/drivers/infiniband/hw/hfi1/user_sdma.h b/drivers/infiniband/hw/hfi1/user_sdma.h index 9972e0e6545e..1e8c02fe8ad1 100644 --- a/drivers/infiniband/hw/hfi1/user_sdma.h +++ b/drivers/infiniband/hw/hfi1/user_sdma.h @@ -1,6 +1,7 @@ #ifndef _HFI1_USER_SDMA_H #define _HFI1_USER_SDMA_H /* + * Copyright(c) 2020 - Cornelis Networks, Inc. * Copyright(c) 2015 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or @@ -133,7 +134,6 @@ struct hfi1_user_sdma_pkt_q { unsigned long unpinned; struct mmu_rb_handler *handler; atomic_t n_locked; - struct mm_struct *mm; }; struct hfi1_user_sdma_comp_q { @@ -250,4 +250,9 @@ int hfi1_user_sdma_process_request(struct hfi1_filedata *fd, struct iovec *iovec, unsigned long dim, unsigned long *count); +static inline struct mm_struct *mm_from_sdma_node(struct sdma_mmu_node *node) +{ + return node->rb.handler->mn.mm; +} + #endif /* _HFI1_USER_SDMA_H */

4 years

1
0
0 0

[PATCH v2 3/5] memory: renesas-rpc-if: Fix a reference leak in rpcif_probe()

by Lad Prabhakar

Release the node reference by calling of_node_put(flash) in the probe. Fixes: ca7d8b980b67f ("memory: add Renesas RPC-IF driver") Reported-by: Pavel Machek <pavel(a)denx.de> Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj(a)bp.renesas.com> Cc: stable(a)vger.kernel.org Reviewed-by: Sergei Shtylyov <sergei.shtylyov(a)gmail.com> --- drivers/memory/renesas-rpc-if.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/memory/renesas-rpc-if.c b/drivers/memory/renesas-rpc-if.c index a8d0ba368625..da0fdb4c7595 100644 --- a/drivers/memory/renesas-rpc-if.c +++ b/drivers/memory/renesas-rpc-if.c @@ -561,9 +561,11 @@ static int rpcif_probe(struct platform_device *pdev) } else if (of_device_is_compatible(flash, "cfi-flash")) { name = "rpc-if-hyperflash"; } else { + of_node_put(flash); dev_warn(&pdev->dev, "unknown flash type\n"); return -ENODEV; } + of_node_put(flash); vdev = platform_device_alloc(name, pdev->id); if (!vdev) -- 2.25.1

4 years

4
3
0 0

[PATCH v2 2/5] memory: renesas-rpc-if: Fix unbalanced pm_runtime_enable in rpcif_{enable, disable}_rpm

by Lad Prabhakar

rpcif_enable_rpm calls pm_runtime_enable, so rpcif_disable_rpm needs to call pm_runtime_disable and not pm_runtime_put_sync. Fixes: ca7d8b980b67f ("memory: add Renesas RPC-IF driver") Reported-by: Reported-by: Geert Uytterhoeven <geert+renesas(a)glider.be> Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj(a)bp.renesas.com> Cc: stable(a)vger.kernel.org --- drivers/memory/renesas-rpc-if.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/memory/renesas-rpc-if.c b/drivers/memory/renesas-rpc-if.c index 69f2e2b4cd50..a8d0ba368625 100644 --- a/drivers/memory/renesas-rpc-if.c +++ b/drivers/memory/renesas-rpc-if.c @@ -212,7 +212,7 @@ EXPORT_SYMBOL(rpcif_enable_rpm); void rpcif_disable_rpm(struct rpcif *rpc) { - pm_runtime_put_sync(rpc->dev); + pm_runtime_disable(rpc->dev); } EXPORT_SYMBOL(rpcif_disable_rpm); -- 2.25.1

4 years

3
2
0 0

[PATCH v2 1/5] memory: renesas-rpc-if: Return correct value to the caller of rpcif_manual_xfer()

by Lad Prabhakar

In the error path of rpcif_manual_xfer() the value of ret is overwritten by value returned by reset_control_reset() function and thus returning incorrect value to the caller. This patch makes sure the correct value is returned to the caller of rpcif_manual_xfer() by dropping the overwrite of ret in error path. Also now we ignore the value returned by reset_control_reset() in the error path and instead print a error message when it fails. Fixes: ca7d8b980b67f ("memory: add Renesas RPC-IF driver") Reported-by: Pavel Machek <pavel(a)denx.de> Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj(a)bp.renesas.com> Cc: stable(a)vger.kernel.org Reviewed-by: Sergei Shtylyov <sergei.shtylyov(a)gmail.com> --- drivers/memory/renesas-rpc-if.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/memory/renesas-rpc-if.c b/drivers/memory/renesas-rpc-if.c index f2a33a1af836..69f2e2b4cd50 100644 --- a/drivers/memory/renesas-rpc-if.c +++ b/drivers/memory/renesas-rpc-if.c @@ -508,7 +508,8 @@ int rpcif_manual_xfer(struct rpcif *rpc) return ret; err_out: - ret = reset_control_reset(rpc->rstc); + if (reset_control_reset(rpc->rstc)) + dev_err(rpc->dev, "Failed to reset HW\n"); rpcif_hw_init(rpc, rpc->bus_size == 2); goto exit; } -- 2.25.1

4 years

4
3
0 0

Re: [Patch stable] netfilter: clear skb->next in NF_HOOK_LIST()

by Greg Kroah-Hartman

On Sat, Nov 28, 2020 at 05:09:18PM +0800, 刘志旭 wrote: > I still didn't see this patch in stable queue yet. Since we've a working POC to panic the  > system (see https://bugzilla.kernel.org/show_bug.cgi?id=209823), I think it's necessary > to merge this patch ASAP, thanks. Odd, I don't think Sasha pushed out his patch queue. I've applied this now, thanks. greg k-h

4 years

1
0
0 0

Re: [PATCH] exit: fix a race in release_task when flushing the dentry

by Greg Kroah-Hartman

On Sat, Nov 28, 2020 at 02:47:22PM +0800, Wen Yang wrote: > [ Upstream commit 7bc3e6e55acf065500a24621f3b313e7e5998acf ] No, that is not this commit at all. What are you wanting to have happen here? confused, greg k-h

4 years

1
0
0 0

[PATCH] mm: memcg/slab: fix obj_cgroup_charge() return value handling

by Roman Gushchin

Commit 10befea91b61 ("mm: memcg/slab: use a single set of kmem_caches for all allocations") introduced a regression into the handling of the obj_cgroup_charge() return value. If a non-zero value is returned (indicating of exceeding one of memory.max limits), the allocation should fail, instead of falling back to non-accounted mode. To make the code more readable, move memcg_slab_pre_alloc_hook() and memcg_slab_post_alloc_hook() calling conditions into bodies of these hooks. Fixes: 10befea91b61 ("mm: memcg/slab: use a single set of kmem_caches for all allocations") Signed-off-by: Roman Gushchin <guro(a)fb.com> Cc: stable(a)vger.kernel.org --- mm/slab.h | 40 ++++++++++++++++++++++++---------------- 1 file changed, 24 insertions(+), 16 deletions(-) diff --git a/mm/slab.h b/mm/slab.h index 59aeb0d9f11b..5dc89d8fb05e 100644 --- a/mm/slab.h +++ b/mm/slab.h @@ -257,22 +257,32 @@ static inline size_t obj_full_size(struct kmem_cache *s) return s->size + sizeof(struct obj_cgroup *); } -static inline struct obj_cgroup *memcg_slab_pre_alloc_hook(struct kmem_cache *s, - size_t objects, - gfp_t flags) +/* + * Returns true if the allocation should fail. + */ +static inline bool memcg_slab_pre_alloc_hook(struct kmem_cache *s, + struct obj_cgroup **objcgp, + size_t objects, gfp_t flags) { struct obj_cgroup *objcg; + if (!memcg_kmem_enabled()) + return false; + + if (!(flags & __GFP_ACCOUNT) && !(s->flags & SLAB_ACCOUNT)) + return false; + objcg = get_obj_cgroup_from_current(); if (!objcg) - return NULL; + return false; if (obj_cgroup_charge(objcg, flags, objects * obj_full_size(s))) { obj_cgroup_put(objcg); - return NULL; + return true; } - return objcg; + *objcgp = objcg; + return false; } static inline void mod_objcg_state(struct obj_cgroup *objcg, @@ -298,7 +308,7 @@ static inline void memcg_slab_post_alloc_hook(struct kmem_cache *s, unsigned long off; size_t i; - if (!objcg) + if (!memcg_kmem_enabled() || !objcg) return; flags &= ~__GFP_ACCOUNT; @@ -382,11 +392,11 @@ static inline void memcg_free_page_obj_cgroups(struct page *page) { } -static inline struct obj_cgroup *memcg_slab_pre_alloc_hook(struct kmem_cache *s, - size_t objects, - gfp_t flags) +static inline bool memcg_slab_pre_alloc_hook(struct kmem_cache *s, + struct obj_cgroup **objcgp, + size_t objects, gfp_t flags) { - return NULL; + return false; } static inline void memcg_slab_post_alloc_hook(struct kmem_cache *s, @@ -494,9 +504,8 @@ static inline struct kmem_cache *slab_pre_alloc_hook(struct kmem_cache *s, if (should_failslab(s, flags)) return NULL; - if (memcg_kmem_enabled() && - ((flags & __GFP_ACCOUNT) || (s->flags & SLAB_ACCOUNT))) - *objcgp = memcg_slab_pre_alloc_hook(s, size, flags); + if (memcg_slab_pre_alloc_hook(s, objcgp, size, flags)) + return NULL; return s; } @@ -515,8 +524,7 @@ static inline void slab_post_alloc_hook(struct kmem_cache *s, s->flags, flags); } - if (memcg_kmem_enabled()) - memcg_slab_post_alloc_hook(s, objcg, flags, size, p); + memcg_slab_post_alloc_hook(s, objcg, flags, size, p); } #ifndef CONFIG_SLOB -- 2.26.2

4 years

2
3
0 0

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror November 2020