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

…el/git/roland/infiniband

Pull infiniband updates from Roland Dreier:
 "Main batch of InfiniBand/RDMA changes for 3.19:

   - On-demand paging support in core midlayer and mlx5 driver.  This
     lets userspace create non-pinned memory regions and have the
     adapter HW trigger page faults.
   - iSER and IPoIB updates and fixes.
   - Low-level HW driver updates for cxgb4, mlx4 and ocrdma.
   - Other miscellaneous fixes"

* tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (56 commits)
  IB/mlx5: Implement on demand paging by adding support for MMU notifiers
  IB/mlx5: Add support for RDMA read/write responder page faults
  IB/mlx5: Handle page faults
  IB/mlx5: Page faults handling infrastructure
  IB/mlx5: Add mlx5_ib_update_mtt to update page tables after creation
  IB/mlx5: Changes in memory region creation to support on-demand paging
  IB/mlx5: Implement the ODP capability query verb
  mlx5_core: Add support for page faults events and low level handling
  mlx5_core: Re-add MLX5_DEV_CAP_FLAG_ON_DMND_PG flag
  IB/srp: Allow newline separator for connection string
  IB/core: Implement support for MMU notifiers regarding on demand paging regions
  IB/core: Add support for on demand paging regions
  IB/core: Add flags for on demand paging support
  IB/core: Add support for extended query device caps
  IB/mlx5: Add function to read WQE from user-space
  IB/core: Add umem function to read data from user-space
  IB/core: Replace ib_umem's offset field with a full address
  IB/mlx5: Enhance UMR support to allow partial page table update
  IB/mlx5: Remove per-MR pas and dma pointers
  RDMA/ocrdma: Always resolve destination mac from GRH for UD QPs
  ...

drivers/infiniband/Kconfig

@@ -38,6 +38,17 @@ config INFINIBAND_USER_MEM
 	depends on INFINIBAND_USER_ACCESS != n
 	default y
 
+config INFINIBAND_ON_DEMAND_PAGING
+	bool "InfiniBand on-demand paging support"
+	depends on INFINIBAND_USER_MEM
+	select MMU_NOTIFIER
+	default y
+	---help---
+	  On demand paging support for the InfiniBand subsystem.
+	  Together with driver support this allows registration of
+	  memory regions without pinning their pages, fetching the
+	  pages on demand instead.
+
 config INFINIBAND_ADDR_TRANS
 	bool
 	depends on INFINIBAND

drivers/infiniband/core/Makefile

@@ -11,6 +11,7 @@ obj-$(CONFIG_INFINIBAND_USER_ACCESS) +=	ib_uverbs.o ib_ucm.o \
 ib_core-y :=			packer.o ud_header.o verbs.o sysfs.o \
 				device.o fmr_pool.o cache.o netlink.o
 ib_core-$(CONFIG_INFINIBAND_USER_MEM) += umem.o
+ib_core-$(CONFIG_INFINIBAND_ON_DEMAND_PAGING) += umem_odp.o umem_rbtree.o
 
 ib_mad-y :=			mad.o smi.o agent.o mad_rmpp.o
 

drivers/infiniband/core/addr.c

@@ -176,8 +176,8 @@ static void set_timeout(unsigned long time)
 	unsigned long delay;
 
 	delay = time - jiffies;
-	if ((long)delay <= 0)
-		delay = 1;
+	if ((long)delay < 0)
+		delay = 0;
 
 	mod_delayed_work(addr_wq, &work, delay);
 }

drivers/infiniband/core/multicast.c

@@ -525,17 +525,22 @@ static void join_handler(int status, struct ib_sa_mcmember_rec *rec,
 	if (status)
 		process_join_error(group, status);
 	else {
+		int mgids_changed, is_mgid0;
 		ib_find_pkey(group->port->dev->device, group->port->port_num,
 			     be16_to_cpu(rec->pkey), &pkey_index);
 
 		spin_lock_irq(&group->port->lock);
-		group->rec = *rec;
 		if (group->state == MCAST_BUSY &&
 		    group->pkey_index == MCAST_INVALID_PKEY_INDEX)
 			group->pkey_index = pkey_index;
-		if (!memcmp(&mgid0, &group->rec.mgid, sizeof mgid0)) {
+		mgids_changed = memcmp(&rec->mgid, &group->rec.mgid,
+				       sizeof(group->rec.mgid));
+		group->rec = *rec;
+		if (mgids_changed) {
 			rb_erase(&group->node, &group->port->table);
-			mcast_insert(group->port, group, 1);
+			is_mgid0 = !memcmp(&mgid0, &group->rec.mgid,
+					   sizeof(mgid0));
+			mcast_insert(group->port, group, is_mgid0);
 		}
 		spin_unlock_irq(&group->port->lock);
 	}

drivers/infiniband/core/umem.c

@@ -39,6 +39,7 @@
 #include <linux/hugetlb.h>
 #include <linux/dma-attrs.h>
 #include <linux/slab.h>
+#include <rdma/ib_umem_odp.h>
 
 #include "uverbs.h"
 
@@ -69,6 +70,10 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
 
 /**
  * ib_umem_get - Pin and DMA map userspace memory.
+ *
+ * If access flags indicate ODP memory, avoid pinning. Instead, stores
+ * the mm for future page fault handling in conjunction with MMU notifiers.
+ *
  * @context: userspace context to pin memory for
  * @addr: userspace virtual address to start at
  * @size: length of region to pin
@@ -103,17 +108,30 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 
 	umem->context   = context;
 	umem->length    = size;
-	umem->offset    = addr & ~PAGE_MASK;
+	umem->address   = addr;
 	umem->page_size = PAGE_SIZE;
 	umem->pid       = get_task_pid(current, PIDTYPE_PID);
 	/*
-	 * We ask for writable memory if any access flags other than
-	 * "remote read" are set.  "Local write" and "remote write"
+	 * We ask for writable memory if any of the following
+	 * access flags are set.  "Local write" and "remote write"
 	 * obviously require write access.  "Remote atomic" can do
 	 * things like fetch and add, which will modify memory, and
 	 * "MW bind" can change permissions by binding a window.
 	 */
-	umem->writable  = !!(access & ~IB_ACCESS_REMOTE_READ);
+	umem->writable  = !!(access &
+		(IB_ACCESS_LOCAL_WRITE   | IB_ACCESS_REMOTE_WRITE |
+		 IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_MW_BIND));
+
+	if (access & IB_ACCESS_ON_DEMAND) {
+		ret = ib_umem_odp_get(context, umem);
+		if (ret) {
+			kfree(umem);
+			return ERR_PTR(ret);
+		}
+		return umem;
+	}
+
+	umem->odp_data = NULL;
 
 	/* We assume the memory is from hugetlb until proved otherwise */
 	umem->hugetlb   = 1;
@@ -132,7 +150,7 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 	if (!vma_list)
 		umem->hugetlb = 0;
 
-	npages = PAGE_ALIGN(size + umem->offset) >> PAGE_SHIFT;
+	npages = ib_umem_num_pages(umem);
 
 	down_write(&current->mm->mmap_sem);
 
@@ -235,6 +253,11 @@ void ib_umem_release(struct ib_umem *umem)
 	struct task_struct *task;
 	unsigned long diff;
 
+	if (umem->odp_data) {
+		ib_umem_odp_release(umem);
+		return;
+	}
+
 	__ib_umem_release(umem->context->device, umem, 1);
 
 	task = get_pid_task(umem->pid, PIDTYPE_PID);
@@ -246,7 +269,7 @@ void ib_umem_release(struct ib_umem *umem)
 	if (!mm)
 		goto out;
 
-	diff = PAGE_ALIGN(umem->length + umem->offset) >> PAGE_SHIFT;
+	diff = ib_umem_num_pages(umem);
 
 	/*
 	 * We may be called with the mm's mmap_sem already held.  This
@@ -283,6 +306,9 @@ int ib_umem_page_count(struct ib_umem *umem)
 	int n;
 	struct scatterlist *sg;
 
+	if (umem->odp_data)
+		return ib_umem_num_pages(umem);
+
 	shift = ilog2(umem->page_size);
 
 	n = 0;
@@ -292,3 +318,37 @@ int ib_umem_page_count(struct ib_umem *umem)
 	return n;
 }
 EXPORT_SYMBOL(ib_umem_page_count);
+
+/*
+ * Copy from the given ib_umem's pages to the given buffer.
+ *
+ * umem - the umem to copy from
+ * offset - offset to start copying from
+ * dst - destination buffer
+ * length - buffer length
+ *
+ * Returns 0 on success, or an error code.
+ */
+int ib_umem_copy_from(void *dst, struct ib_umem *umem, size_t offset,
+		      size_t length)
+{
+	size_t end = offset + length;
+	int ret;
+
+	if (offset > umem->length || length > umem->length - offset) {
+		pr_err("ib_umem_copy_from not in range. offset: %zd umem length: %zd end: %zd\n",
+		       offset, umem->length, end);
+		return -EINVAL;
+	}
+
+	ret = sg_pcopy_to_buffer(umem->sg_head.sgl, umem->nmap, dst, length,
+				 offset + ib_umem_offset(umem));
+
+	if (ret < 0)
+		return ret;
+	else if (ret != length)
+		return -EINVAL;
+	else
+		return 0;
+}
+EXPORT_SYMBOL(ib_umem_copy_from);