IB/mlx5: Refactoring register MR code
In order to add re-registration of memory region, some logic was
extracted to separate functions:
- ODP related logic.
- Some of the UMR WQE preparation code.
- DMA mapping.
- Umem creation.
- Creating MKey using FW interface.
- MR fields assignments after successful creation.
Signed-off-by: Noa Osherovich <noaos@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 6000f7a..9d6dade 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -77,6 +77,34 @@
return order - cache->ent[0].order;
}
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+static void update_odp_mr(struct mlx5_ib_mr *mr)
+{
+ if (mr->umem->odp_data) {
+ /*
+ * This barrier prevents the compiler from moving the
+ * setting of umem->odp_data->private to point to our
+ * MR, before reg_umr finished, to ensure that the MR
+ * initialization have finished before starting to
+ * handle invalidations.
+ */
+ smp_wmb();
+ mr->umem->odp_data->private = mr;
+ /*
+ * Make sure we will see the new
+ * umem->odp_data->private value in the invalidation
+ * routines, before we can get page faults on the
+ * MR. Page faults can happen once we put the MR in
+ * the tree, below this line. Without the barrier,
+ * there can be a fault handling and an invalidation
+ * before umem->odp_data->private == mr is visible to
+ * the invalidation handler.
+ */
+ smp_wmb();
+ }
+}
+#endif
+
static void reg_mr_callback(int status, void *context)
{
struct mlx5_ib_mr *mr = context;
@@ -693,10 +721,40 @@
return order <= MLX5_MAX_UMR_SHIFT;
}
-static void prep_umr_reg_wqe(struct ib_pd *pd, struct ib_send_wr *wr,
- struct ib_sge *sg, u64 dma, int n, u32 key,
- int page_shift, u64 virt_addr, u64 len,
- int access_flags)
+static int dma_map_mr_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
+ int npages, int page_shift, int *size,
+ __be64 **mr_pas, dma_addr_t *dma)
+{
+ __be64 *pas;
+ struct device *ddev = dev->ib_dev.dma_device;
+
+ /*
+ * UMR copies MTTs in units of MLX5_UMR_MTT_ALIGNMENT bytes.
+ * To avoid copying garbage after the pas array, we allocate
+ * a little more.
+ */
+ *size = ALIGN(sizeof(u64) * npages, MLX5_UMR_MTT_ALIGNMENT);
+ *mr_pas = kmalloc(*size + MLX5_UMR_ALIGN - 1, GFP_KERNEL);
+ if (!(*mr_pas))
+ return -ENOMEM;
+
+ pas = PTR_ALIGN(*mr_pas, MLX5_UMR_ALIGN);
+ mlx5_ib_populate_pas(dev, umem, page_shift, pas, MLX5_IB_MTT_PRESENT);
+ /* Clear padding after the actual pages. */
+ memset(pas + npages, 0, *size - npages * sizeof(u64));
+
+ *dma = dma_map_single(ddev, pas, *size, DMA_TO_DEVICE);
+ if (dma_mapping_error(ddev, *dma)) {
+ kfree(*mr_pas);
+ return -ENOMEM;
+ }
+
+ return 0;
+}
+
+static void prep_umr_wqe_common(struct ib_pd *pd, struct ib_send_wr *wr,
+ struct ib_sge *sg, u64 dma, int n, u32 key,
+ int page_shift)
{
struct mlx5_ib_dev *dev = to_mdev(pd->device);
struct mlx5_umr_wr *umrwr = umr_wr(wr);
@@ -706,7 +764,6 @@
sg->lkey = dev->umrc.pd->local_dma_lkey;
wr->next = NULL;
- wr->send_flags = 0;
wr->sg_list = sg;
if (n)
wr->num_sge = 1;
@@ -718,6 +775,19 @@
umrwr->npages = n;
umrwr->page_shift = page_shift;
umrwr->mkey = key;
+}
+
+static void prep_umr_reg_wqe(struct ib_pd *pd, struct ib_send_wr *wr,
+ struct ib_sge *sg, u64 dma, int n, u32 key,
+ int page_shift, u64 virt_addr, u64 len,
+ int access_flags)
+{
+ struct mlx5_umr_wr *umrwr = umr_wr(wr);
+
+ prep_umr_wqe_common(pd, wr, sg, dma, n, key, page_shift);
+
+ wr->send_flags = 0;
+
umrwr->target.virt_addr = virt_addr;
umrwr->length = len;
umrwr->access_flags = access_flags;
@@ -734,6 +804,31 @@
umrwr->mkey = key;
}
+static struct ib_umem *mr_umem_get(struct ib_pd *pd, u64 start, u64 length,
+ int access_flags, int *npages,
+ int *page_shift, int *ncont, int *order)
+{
+ struct mlx5_ib_dev *dev = to_mdev(pd->device);
+ struct ib_umem *umem = ib_umem_get(pd->uobject->context, start, length,
+ access_flags, 0);
+ if (IS_ERR(umem)) {
+ mlx5_ib_err(dev, "umem get failed (%ld)\n", PTR_ERR(umem));
+ return (void *)umem;
+ }
+
+ mlx5_ib_cont_pages(umem, start, npages, page_shift, ncont, order);
+ if (!*npages) {
+ mlx5_ib_warn(dev, "avoid zero region\n");
+ ib_umem_release(umem);
+ return ERR_PTR(-EINVAL);
+ }
+
+ mlx5_ib_dbg(dev, "npages %d, ncont %d, order %d, page_shift %d\n",
+ *npages, *ncont, *order, *page_shift);
+
+ return umem;
+}
+
void mlx5_umr_cq_handler(struct ib_cq *cq, void *cq_context)
{
struct mlx5_ib_umr_context *context;
@@ -770,7 +865,6 @@
struct ib_sge sg;
int size;
__be64 *mr_pas;
- __be64 *pas;
dma_addr_t dma;
int err = 0;
int i;
@@ -790,26 +884,10 @@
if (!mr)
return ERR_PTR(-EAGAIN);
- /* UMR copies MTTs in units of MLX5_UMR_MTT_ALIGNMENT bytes.
- * To avoid copying garbage after the pas array, we allocate
- * a little more. */
- size = ALIGN(sizeof(u64) * npages, MLX5_UMR_MTT_ALIGNMENT);
- mr_pas = kmalloc(size + MLX5_UMR_ALIGN - 1, GFP_KERNEL);
- if (!mr_pas) {
- err = -ENOMEM;
+ err = dma_map_mr_pas(dev, umem, npages, page_shift, &size, &mr_pas,
+ &dma);
+ if (err)
goto free_mr;
- }
-
- pas = PTR_ALIGN(mr_pas, MLX5_UMR_ALIGN);
- mlx5_ib_populate_pas(dev, umem, page_shift, pas, MLX5_IB_MTT_PRESENT);
- /* Clear padding after the actual pages. */
- memset(pas + npages, 0, size - npages * sizeof(u64));
-
- dma = dma_map_single(ddev, pas, size, DMA_TO_DEVICE);
- if (dma_mapping_error(ddev, dma)) {
- err = -ENOMEM;
- goto free_pas;
- }
memset(&umrwr, 0, sizeof(umrwr));
umrwr.wr.wr_id = (u64)(unsigned long)&umr_context;
@@ -840,7 +918,6 @@
up(&umrc->sem);
dma_unmap_single(ddev, dma, size, DMA_TO_DEVICE);
-free_pas:
kfree(mr_pas);
free_mr:
@@ -974,10 +1051,14 @@
}
#endif
-static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, u64 virt_addr,
- u64 length, struct ib_umem *umem,
- int npages, int page_shift,
- int access_flags)
+/*
+ * If ibmr is NULL it will be allocated by reg_create.
+ * Else, the given ibmr will be used.
+ */
+static struct mlx5_ib_mr *reg_create(struct ib_mr *ibmr, struct ib_pd *pd,
+ u64 virt_addr, u64 length,
+ struct ib_umem *umem, int npages,
+ int page_shift, int access_flags)
{
struct mlx5_ib_dev *dev = to_mdev(pd->device);
struct mlx5_create_mkey_mbox_in *in;
@@ -986,7 +1067,7 @@
int err;
bool pg_cap = !!(MLX5_CAP_GEN(dev->mdev, pg));
- mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+ mr = ibmr ? to_mmr(ibmr) : kzalloc(sizeof(*mr), GFP_KERNEL);
if (!mr)
return ERR_PTR(-ENOMEM);
@@ -1032,11 +1113,22 @@
kvfree(in);
err_1:
- kfree(mr);
+ if (!ibmr)
+ kfree(mr);
return ERR_PTR(err);
}
+static void set_mr_fileds(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
+ int npages, u64 length, int access_flags)
+{
+ mr->npages = npages;
+ atomic_add(npages, &dev->mdev->priv.reg_pages);
+ mr->ibmr.lkey = mr->mmr.key;
+ mr->ibmr.rkey = mr->mmr.key;
+ mr->ibmr.length = length;
+}
+
struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
u64 virt_addr, int access_flags,
struct ib_udata *udata)
@@ -1052,22 +1144,11 @@
mlx5_ib_dbg(dev, "start 0x%llx, virt_addr 0x%llx, length 0x%llx, access_flags 0x%x\n",
start, virt_addr, length, access_flags);
- umem = ib_umem_get(pd->uobject->context, start, length, access_flags,
- 0);
- if (IS_ERR(umem)) {
- mlx5_ib_dbg(dev, "umem get failed (%ld)\n", PTR_ERR(umem));
+ umem = mr_umem_get(pd, start, length, access_flags, &npages,
+ &page_shift, &ncont, &order);
+
+ if (IS_ERR(umem))
return (void *)umem;
- }
-
- mlx5_ib_cont_pages(umem, start, &npages, &page_shift, &ncont, &order);
- if (!npages) {
- mlx5_ib_warn(dev, "avoid zero region\n");
- err = -EINVAL;
- goto error;
- }
-
- mlx5_ib_dbg(dev, "npages %d, ncont %d, order %d, page_shift %d\n",
- npages, ncont, order, page_shift);
if (use_umr(order)) {
mr = reg_umr(pd, umem, virt_addr, length, ncont, page_shift,
@@ -1083,8 +1164,8 @@
}
if (!mr)
- mr = reg_create(pd, virt_addr, length, umem, ncont, page_shift,
- access_flags);
+ mr = reg_create(NULL, pd, virt_addr, length, umem, ncont,
+ page_shift, access_flags);
if (IS_ERR(mr)) {
err = PTR_ERR(mr);
@@ -1094,34 +1175,10 @@
mlx5_ib_dbg(dev, "mkey 0x%x\n", mr->mmr.key);
mr->umem = umem;
- mr->npages = npages;
- atomic_add(npages, &dev->mdev->priv.reg_pages);
- mr->ibmr.lkey = mr->mmr.key;
- mr->ibmr.rkey = mr->mmr.key;
+ set_mr_fileds(dev, mr, npages, length, access_flags);
#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
- if (umem->odp_data) {
- /*
- * This barrier prevents the compiler from moving the
- * setting of umem->odp_data->private to point to our
- * MR, before reg_umr finished, to ensure that the MR
- * initialization have finished before starting to
- * handle invalidations.
- */
- smp_wmb();
- mr->umem->odp_data->private = mr;
- /*
- * Make sure we will see the new
- * umem->odp_data->private value in the invalidation
- * routines, before we can get page faults on the
- * MR. Page faults can happen once we put the MR in
- * the tree, below this line. Without the barrier,
- * there can be a fault handling and an invalidation
- * before umem->odp_data->private == mr is visible to
- * the invalidation handler.
- */
- smp_wmb();
- }
+ update_odp_mr(mr);
#endif
return &mr->ibmr;