Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "Two security issues and several small bug fixes. Things seem to have
  stabilized for this release here.

  Summary:

   - Significant out of bounds access security issue in i40iw

   - Fix misuse of mmu notifiers in hfi1

   - Several errors in the register map/usage in hns

   - Missing error returns in mthca"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/hns: Bugfix for memory window mtpt configuration
  RDMA/hns: Fix retry_cnt and rnr_cnt when querying QP
  RDMA/hns: Fix wrong field of SRQ number the device supports
  IB/hfi1: Ensure correct mm is used at all times
  RDMA/i40iw: Address an mmap handler exploit in i40iw
  IB/mthca: fix return value of error branch in mthca_init_cq()

drivers/infiniband/hw/hfi1/file_ops.c

 /*
+ * 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);

drivers/infiniband/hw/hfi1/hfi.h

 #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;

drivers/infiniband/hw/hfi1/mmu_rb.c

 /*
+ * 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);

drivers/infiniband/hw/hfi1/mmu_rb.h

 /*
+ * 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);

drivers/infiniband/hw/hfi1/user_exp_rcv.c

 /*
+ * 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)

drivers/infiniband/hw/hfi1/user_exp_rcv.h

 #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 */

drivers/infiniband/hw/hfi1/user_sdma.c

 /*
+ * 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);
 	}
 }

drivers/infiniband/hw/hfi1/user_sdma.h

 #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 */

drivers/infiniband/hw/hns/hns_roce_hw_v2.c

@@ -2936,6 +2936,7 @@ static int hns_roce_v2_mw_write_mtpt(void *mb_buf, struct hns_roce_mw *mw)
 
 	roce_set_bit(mpt_entry->byte_8_mw_cnt_en, V2_MPT_BYTE_8_R_INV_EN_S, 1);
 	roce_set_bit(mpt_entry->byte_8_mw_cnt_en, V2_MPT_BYTE_8_L_INV_EN_S, 1);
+	roce_set_bit(mpt_entry->byte_8_mw_cnt_en, V2_MPT_BYTE_8_LW_EN_S, 1);
 
 	roce_set_bit(mpt_entry->byte_12_mw_pa, V2_MPT_BYTE_12_PA_S, 0);
 	roce_set_bit(mpt_entry->byte_12_mw_pa, V2_MPT_BYTE_12_MR_MW_S, 1);
@@ -4989,11 +4990,11 @@ static int hns_roce_v2_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 					      V2_QPC_BYTE_28_AT_M,
 					      V2_QPC_BYTE_28_AT_S);
 	qp_attr->retry_cnt = roce_get_field(context.byte_212_lsn,
-					    V2_QPC_BYTE_212_RETRY_CNT_M,
-					    V2_QPC_BYTE_212_RETRY_CNT_S);
+					    V2_QPC_BYTE_212_RETRY_NUM_INIT_M,
+					    V2_QPC_BYTE_212_RETRY_NUM_INIT_S);
 	qp_attr->rnr_retry = roce_get_field(context.byte_244_rnr_rxack,
-					    V2_QPC_BYTE_244_RNR_CNT_M,
-					    V2_QPC_BYTE_244_RNR_CNT_S);
+					    V2_QPC_BYTE_244_RNR_NUM_INIT_M,
+					    V2_QPC_BYTE_244_RNR_NUM_INIT_S);
 
 done:
 	qp_attr->cur_qp_state = qp_attr->qp_state;

drivers/infiniband/hw/hns/hns_roce_hw_v2.h

@@ -1661,7 +1661,7 @@ struct hns_roce_query_pf_caps_d {
 	__le32 rsv_uars_rsv_qps;
 };
 #define V2_QUERY_PF_CAPS_D_NUM_SRQS_S 0
-#define V2_QUERY_PF_CAPS_D_NUM_SRQS_M GENMASK(20, 0)
+#define V2_QUERY_PF_CAPS_D_NUM_SRQS_M GENMASK(19, 0)
 
 #define V2_QUERY_PF_CAPS_D_RQWQE_HOP_NUM_S 20
 #define V2_QUERY_PF_CAPS_D_RQWQE_HOP_NUM_M GENMASK(21, 20)

drivers/infiniband/hw/i40iw/i40iw_main.c

@@ -54,10 +54,6 @@
 #define DRV_VERSION	__stringify(DRV_VERSION_MAJOR) "."		\
 	__stringify(DRV_VERSION_MINOR) "." __stringify(DRV_VERSION_BUILD)
 
-static int push_mode;
-module_param(push_mode, int, 0644);
-MODULE_PARM_DESC(push_mode, "Low latency mode: 0=disabled (default), 1=enabled)");
-
 static int debug;
 module_param(debug, int, 0644);
 MODULE_PARM_DESC(debug, "debug flags: 0=disabled (default), 0x7fffffff=all");
@@ -1580,7 +1576,6 @@ static enum i40iw_status_code i40iw_setup_init_state(struct i40iw_handler *hdl,
 	if (status)
 		goto exit;
 	iwdev->obj_next = iwdev->obj_mem;
-	iwdev->push_mode = push_mode;
 
 	init_waitqueue_head(&iwdev->vchnl_waitq);
 	init_waitqueue_head(&dev->vf_reqs);

drivers/infiniband/hw/i40iw/i40iw_verbs.c

@@ -167,39 +167,16 @@ static void i40iw_dealloc_ucontext(struct ib_ucontext *context)
  */
 static int i40iw_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
 {
-	struct i40iw_ucontext *ucontext;
-	u64 db_addr_offset, push_offset, pfn;
-
-	ucontext = to_ucontext(context);
-	if (ucontext->iwdev->sc_dev.is_pf) {
-		db_addr_offset = I40IW_DB_ADDR_OFFSET;
-		push_offset = I40IW_PUSH_OFFSET;
-		if (vma->vm_pgoff)
-			vma->vm_pgoff += I40IW_PF_FIRST_PUSH_PAGE_INDEX - 1;
-	} else {
-		db_addr_offset = I40IW_VF_DB_ADDR_OFFSET;
-		push_offset = I40IW_VF_PUSH_OFFSET;
-		if (vma->vm_pgoff)
-			vma->vm_pgoff += I40IW_VF_FIRST_PUSH_PAGE_INDEX - 1;
-	}
+	struct i40iw_ucontext *ucontext = to_ucontext(context);
+	u64 dbaddr;
 
-	vma->vm_pgoff += db_addr_offset >> PAGE_SHIFT;
-
-	if (vma->vm_pgoff == (db_addr_offset >> PAGE_SHIFT)) {
-		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
-	} else {
-		if ((vma->vm_pgoff - (push_offset >> PAGE_SHIFT)) % 2)
-			vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
-		else
-			vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
-	}
+	if (vma->vm_pgoff || vma->vm_end - vma->vm_start != PAGE_SIZE)
+		return -EINVAL;
 
-	pfn = vma->vm_pgoff +
-	      (pci_resource_start(ucontext->iwdev->ldev->pcidev, 0) >>
-	       PAGE_SHIFT);
+	dbaddr = I40IW_DB_ADDR_OFFSET + pci_resource_start(ucontext->iwdev->ldev->pcidev, 0);
 
-	return rdma_user_mmap_io(context, vma, pfn, PAGE_SIZE,
-				 vma->vm_page_prot, NULL);
+	return rdma_user_mmap_io(context, vma, dbaddr >> PAGE_SHIFT, PAGE_SIZE,
+				 pgprot_noncached(vma->vm_page_prot), NULL);
 }
 
 /**

drivers/infiniband/hw/mthca/mthca_cq.c

@@ -803,8 +803,10 @@ int mthca_init_cq(struct mthca_dev *dev, int nent,
 	}
 
 	mailbox = mthca_alloc_mailbox(dev, GFP_KERNEL);
-	if (IS_ERR(mailbox))
+	if (IS_ERR(mailbox)) {
+		err = PTR_ERR(mailbox);
 		goto err_out_arm;
+	}
 
 	cq_context = mailbox->buf;
 
@@ -846,9 +848,9 @@ int mthca_init_cq(struct mthca_dev *dev, int nent,
 	}
 
 	spin_lock_irq(&dev->cq_table.lock);
-	if (mthca_array_set(&dev->cq_table.cq,
-			    cq->cqn & (dev->limits.num_cqs - 1),
-			    cq)) {
+	err = mthca_array_set(&dev->cq_table.cq,
+			      cq->cqn & (dev->limits.num_cqs - 1), cq);
+	if (err) {
 		spin_unlock_irq(&dev->cq_table.lock);
 		goto err_out_free_mr;
 	}