Signed-off-by: Roman Pen <roman.penyaev@xxxxxxxxxxxxxxxx>
Cc: Christoph Hellwig <hch@xxxxxx>
Cc: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
Cc: Bart Van Assche <bart.vanassche@xxxxxxxxxxx>
Cc: Sagi Grimberg <sagi@xxxxxxxxxxx>
Cc: Doug Ledford <dledford@xxxxxxxxxx>
---
drivers/nvme/target/rdma.c | 180 ++++++++++++++++++---------------------------
1 file changed, 72 insertions(+), 108 deletions(-)
diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index 52e0c5d579a7..0bafdf3456b8 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -27,6 +27,7 @@
#include <rdma/ib_verbs.h>
#include <rdma/rdma_cm.h>
+#include <rdma/dev_pool.h>
#include <rdma/rw.h>
#include <linux/nvme-rdma.h>
@@ -109,13 +110,10 @@ struct nvmet_rdma_queue {
};
struct nvmet_rdma_device {
- struct ib_device *device;
- struct ib_pd *pd;
+ struct ib_pool_device dev;
struct ib_srq *srq;
struct nvmet_rdma_cmd *srq_cmds;
size_t srq_size;
- struct kref ref;
- struct list_head entry;
};
static bool nvmet_rdma_use_srq;
@@ -126,8 +124,7 @@ static DEFINE_IDA(nvmet_rdma_queue_ida);
static LIST_HEAD(nvmet_rdma_queue_list);
static DEFINE_MUTEX(nvmet_rdma_queue_mutex);
-static LIST_HEAD(device_list);
-static DEFINE_MUTEX(device_list_mutex);
+static struct ib_device_pool ib_devs_pool;
static bool nvmet_rdma_execute_command(struct nvmet_rdma_rsp *rsp);
static void nvmet_rdma_send_done(struct ib_cq *cq, struct ib_wc *wc);
@@ -192,26 +189,26 @@ static int nvmet_rdma_alloc_cmd(struct nvmet_rdma_device *ndev,
if (!c->nvme_cmd)
goto out;
- c->sge[0].addr = ib_dma_map_single(ndev->device, c->nvme_cmd,
+ c->sge[0].addr = ib_dma_map_single(ndev->dev.ib_dev, c->nvme_cmd,
sizeof(*c->nvme_cmd), DMA_FROM_DEVICE);
- if (ib_dma_mapping_error(ndev->device, c->sge[0].addr))
+ if (ib_dma_mapping_error(ndev->dev.ib_dev, c->sge[0].addr))
goto out_free_cmd;
c->sge[0].length = sizeof(*c->nvme_cmd);
- c->sge[0].lkey = ndev->pd->local_dma_lkey;
+ c->sge[0].lkey = ndev->dev.ib_pd->local_dma_lkey;
if (!admin) {
c->inline_page = alloc_pages(GFP_KERNEL,
get_order(NVMET_RDMA_INLINE_DATA_SIZE));
if (!c->inline_page)
goto out_unmap_cmd;
- c->sge[1].addr = ib_dma_map_page(ndev->device,
+ c->sge[1].addr = ib_dma_map_page(ndev->dev.ib_dev,
c->inline_page, 0, NVMET_RDMA_INLINE_DATA_SIZE,
DMA_FROM_DEVICE);
- if (ib_dma_mapping_error(ndev->device, c->sge[1].addr))
+ if (ib_dma_mapping_error(ndev->dev.ib_dev, c->sge[1].addr))
goto out_free_inline_page;
c->sge[1].length = NVMET_RDMA_INLINE_DATA_SIZE;
- c->sge[1].lkey = ndev->pd->local_dma_lkey;
+ c->sge[1].lkey = ndev->dev.ib_pd->local_dma_lkey;
}
c->cqe.done = nvmet_rdma_recv_done;
@@ -228,7 +225,7 @@ static int nvmet_rdma_alloc_cmd(struct nvmet_rdma_device *ndev,
get_order(NVMET_RDMA_INLINE_DATA_SIZE));
}
out_unmap_cmd:
- ib_dma_unmap_single(ndev->device, c->sge[0].addr,
+ ib_dma_unmap_single(ndev->dev.ib_dev, c->sge[0].addr,
sizeof(*c->nvme_cmd), DMA_FROM_DEVICE);
out_free_cmd:
kfree(c->nvme_cmd);
@@ -241,12 +238,12 @@ static void nvmet_rdma_free_cmd(struct nvmet_rdma_device *ndev,
struct nvmet_rdma_cmd *c, bool admin)
{
if (!admin) {
- ib_dma_unmap_page(ndev->device, c->sge[1].addr,
+ ib_dma_unmap_page(ndev->dev.ib_dev, c->sge[1].addr,
NVMET_RDMA_INLINE_DATA_SIZE, DMA_FROM_DEVICE);
__free_pages(c->inline_page,
get_order(NVMET_RDMA_INLINE_DATA_SIZE));
}
- ib_dma_unmap_single(ndev->device, c->sge[0].addr,
+ ib_dma_unmap_single(ndev->dev.ib_dev, c->sge[0].addr,
sizeof(*c->nvme_cmd), DMA_FROM_DEVICE);
kfree(c->nvme_cmd);
}
@@ -296,13 +293,13 @@ static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev,
if (!r->req.rsp)
goto out;
- r->send_sge.addr = ib_dma_map_single(ndev->device, r->req.rsp,
+ r->send_sge.addr = ib_dma_map_single(ndev->dev.ib_dev, r->req.rsp,
sizeof(*r->req.rsp), DMA_TO_DEVICE);
- if (ib_dma_mapping_error(ndev->device, r->send_sge.addr))
+ if (ib_dma_mapping_error(ndev->dev.ib_dev, r->send_sge.addr))
goto out_free_rsp;
r->send_sge.length = sizeof(*r->req.rsp);
- r->send_sge.lkey = ndev->pd->local_dma_lkey;
+ r->send_sge.lkey = ndev->dev.ib_pd->local_dma_lkey;
r->send_cqe.done = nvmet_rdma_send_done;
@@ -324,7 +321,7 @@ static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev,
static void nvmet_rdma_free_rsp(struct nvmet_rdma_device *ndev,
struct nvmet_rdma_rsp *r)
{
- ib_dma_unmap_single(ndev->device, r->send_sge.addr,
+ ib_dma_unmap_single(ndev->dev.ib_dev, r->send_sge.addr,
sizeof(*r->req.rsp), DMA_TO_DEVICE);
kfree(r->req.rsp);
}
@@ -384,7 +381,7 @@ static int nvmet_rdma_post_recv(struct nvmet_rdma_device *ndev,
{
struct ib_recv_wr *bad_wr;
- ib_dma_sync_single_for_device(ndev->device,
+ ib_dma_sync_single_for_device(ndev->dev.ib_dev,
cmd->sge[0].addr, cmd->sge[0].length,
DMA_FROM_DEVICE);
@@ -489,7 +486,7 @@ static void nvmet_rdma_queue_response(struct nvmet_req *req)
nvmet_rdma_post_recv(rsp->queue->dev, rsp->cmd);
- ib_dma_sync_single_for_device(rsp->queue->dev->device,
+ ib_dma_sync_single_for_device(rsp->queue->dev->dev.ib_dev,
rsp->send_sge.addr, rsp->send_sge.length,
DMA_TO_DEVICE);
@@ -650,10 +647,10 @@ static void nvmet_rdma_handle_command(struct nvmet_rdma_queue *queue,
{
u16 status;
- ib_dma_sync_single_for_cpu(queue->dev->device,
+ ib_dma_sync_single_for_cpu(queue->dev->dev.ib_dev,
cmd->cmd->sge[0].addr, cmd->cmd->sge[0].length,
DMA_FROM_DEVICE);
- ib_dma_sync_single_for_cpu(queue->dev->device,
+ ib_dma_sync_single_for_cpu(queue->dev->dev.ib_dev,
cmd->send_sge.addr, cmd->send_sge.length,
DMA_TO_DEVICE);
@@ -746,7 +743,7 @@ static int nvmet_rdma_init_srq(struct nvmet_rdma_device *ndev)
srq_attr.attr.max_sge = 2;
srq_attr.attr.srq_limit = 0;
srq_attr.srq_type = IB_SRQT_BASIC;
- srq = ib_create_srq(ndev->pd, &srq_attr);
+ srq = ib_create_srq(ndev->dev.ib_pd, &srq_attr);
if (IS_ERR(srq)) {
/*
* If SRQs aren't supported we just go ahead and use normal
@@ -775,66 +772,6 @@ static int nvmet_rdma_init_srq(struct nvmet_rdma_device *ndev)
return ret;
}
-static void nvmet_rdma_free_dev(struct kref *ref)
-{
- struct nvmet_rdma_device *ndev =
- container_of(ref, struct nvmet_rdma_device, ref);
-
- mutex_lock(&device_list_mutex);
- list_del(&ndev->entry);
- mutex_unlock(&device_list_mutex);
-
- nvmet_rdma_destroy_srq(ndev);
- ib_dealloc_pd(ndev->pd);
-
- kfree(ndev);
-}
-
-static struct nvmet_rdma_device *
-nvmet_rdma_find_get_device(struct rdma_cm_id *cm_id)
-{
- struct nvmet_rdma_device *ndev;
- int ret;
-
- mutex_lock(&device_list_mutex);
- list_for_each_entry(ndev, &device_list, entry) {
- if (ndev->device->node_guid == cm_id->device->node_guid &&
- kref_get_unless_zero(&ndev->ref))
- goto out_unlock;
- }
-
- ndev = kzalloc(sizeof(*ndev), GFP_KERNEL);
- if (!ndev)
- goto out_err;
-
- ndev->device = cm_id->device;
- kref_init(&ndev->ref);
-
- ndev->pd = ib_alloc_pd(ndev->device, 0);
- if (IS_ERR(ndev->pd))
- goto out_free_dev;
-
- if (nvmet_rdma_use_srq) {
- ret = nvmet_rdma_init_srq(ndev);
- if (ret)
- goto out_free_pd;
- }
-
- list_add(&ndev->entry, &device_list);
-out_unlock:
- mutex_unlock(&device_list_mutex);
- pr_debug("added %s.\n", ndev->device->name);
- return ndev;
-
-out_free_pd:
- ib_dealloc_pd(ndev->pd);
-out_free_dev:
- kfree(ndev);
-out_err:
- mutex_unlock(&device_list_mutex);
- return NULL;
-}
-
static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
{
struct ib_qp_init_attr qp_attr;
@@ -846,14 +783,14 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
* but still keep all admin queues on vector 0.
*/
comp_vector = !queue->host_qid ? 0 :
- queue->idx % ndev->device->num_comp_vectors;
+ queue->idx % ndev->dev.ib_dev->num_comp_vectors;
/*
* Reserve CQ slots for RECV + RDMA_READ/RDMA_WRITE + RDMA_SEND.
*/
nr_cqe = queue->recv_queue_size + 2 * queue->send_queue_size;
- queue->cq = ib_alloc_cq(ndev->device, queue,
+ queue->cq = ib_alloc_cq(ndev->dev.ib_dev, queue,
nr_cqe + 1, comp_vector,
IB_POLL_WORKQUEUE);
if (IS_ERR(queue->cq)) {
@@ -873,8 +810,8 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
/* +1 for drain */
qp_attr.cap.max_send_wr = queue->send_queue_size + 1;
qp_attr.cap.max_rdma_ctxs = queue->send_queue_size;
- qp_attr.cap.max_send_sge = max(ndev->device->attrs.max_sge_rd,
- ndev->device->attrs.max_sge);
+ qp_attr.cap.max_send_sge = max(ndev->dev.ib_dev->attrs.max_sge_rd,
+ ndev->dev.ib_dev->attrs.max_sge);
if (ndev->srq) {
qp_attr.srq = ndev->srq;
@@ -884,7 +821,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
qp_attr.cap.max_recv_sge = 2;
}
- ret = rdma_create_qp(queue->cm_id, ndev->pd, &qp_attr);
+ ret = rdma_create_qp(queue->cm_id, ndev->dev.ib_pd, &qp_attr);
if (ret) {
pr_err("failed to create_qp ret= %d\n", ret);
goto err_destroy_cq;
@@ -942,11 +879,10 @@ static void nvmet_rdma_release_queue_work(struct work_struct *w)
{
struct nvmet_rdma_queue *queue =
container_of(w, struct nvmet_rdma_queue, release_work);
- struct nvmet_rdma_device *dev = queue->dev;
+ struct nvmet_rdma_device *ndev = queue->dev;
nvmet_rdma_free_queue(queue);
-
- kref_put(&dev->ref, nvmet_rdma_free_dev);
+ ib_pool_dev_put(&ndev->dev);
}
static int
@@ -1111,7 +1047,7 @@ static int nvmet_rdma_cm_accept(struct rdma_cm_id *cm_id,
param.rnr_retry_count = 7;
param.flow_control = 1;
param.initiator_depth = min_t(u8, p->initiator_depth,
- queue->dev->device->attrs.max_qp_init_rd_atom);
+ queue->dev->dev.ib_dev->attrs.max_qp_init_rd_atom);
param.private_data = &priv;
param.private_data_len = sizeof(priv);
priv.recfmt = cpu_to_le16(NVME_RDMA_CM_FMT_1_0);
@@ -1129,13 +1065,15 @@ static int nvmet_rdma_queue_connect(struct rdma_cm_id *cm_id,
{
struct nvmet_rdma_device *ndev;
struct nvmet_rdma_queue *queue;
+ struct ib_pool_device *pool_dev;
int ret = -EINVAL;
- ndev = nvmet_rdma_find_get_device(cm_id);
- if (!ndev) {
+ pool_dev = ib_pool_dev_find_get_or_create(cm_id->device, &ib_devs_pool);
+ if (unlikely(!pool_dev)) {
nvmet_rdma_cm_reject(cm_id, NVME_RDMA_CM_NO_RSC);
return -ECONNREFUSED;
}
+ ndev = container_of(pool_dev, typeof(*ndev), dev);
queue = nvmet_rdma_alloc_queue(ndev, cm_id, event);
if (!queue) {
@@ -1163,7 +1101,7 @@ static int nvmet_rdma_queue_connect(struct rdma_cm_id *cm_id,
return 0;
put_device:
- kref_put(&ndev->ref, nvmet_rdma_free_dev);
+ ib_pool_dev_put(&ndev->dev);
return ret;
}
@@ -1469,18 +1407,9 @@ static const struct nvmet_fabrics_ops nvmet_rdma_ops = {
static void nvmet_rdma_remove_one(struct ib_device *ib_device, void *client_data)
{
struct nvmet_rdma_queue *queue, *tmp;
- struct nvmet_rdma_device *ndev;
- bool found = false;
-
- mutex_lock(&device_list_mutex);
- list_for_each_entry(ndev, &device_list, entry) {
- if (ndev->device == ib_device) {
- found = true;
- break;
- }
- }
- mutex_unlock(&device_list_mutex);
+ bool found;
+ found = ib_pool_dev_exists(ib_device, &ib_devs_pool);
if (!found)
return;
@@ -1491,7 +1420,7 @@ static void nvmet_rdma_remove_one(struct ib_device *ib_device, void *client_data
mutex_lock(&nvmet_rdma_queue_mutex);
list_for_each_entry_safe(queue, tmp, &nvmet_rdma_queue_list,
queue_list) {
- if (queue->dev->device != ib_device)
+ if (queue->dev->dev.ib_dev != ib_device)
continue;
pr_info("Removing queue %d\n", queue->idx);
@@ -1508,10 +1437,44 @@ static struct ib_client nvmet_rdma_ib_client = {
.remove = nvmet_rdma_remove_one
};
+static struct ib_pool_device *nvmet_ib_pool_dev_alloc(void)
+{
+ struct nvmet_rdma_device *ndev;
+
+ ndev = kzalloc(sizeof(*ndev), GFP_KERNEL);
+ if (likely(ndev))
+ return &ndev->dev;
+
+ return NULL;
+}
+
+static void nvmet_ib_pool_dev_free(struct ib_pool_device *dev)
+{
+ struct nvmet_rdma_device *ndev;
+
+ ndev = container_of(dev, typeof(*ndev), dev);
+ nvmet_rdma_destroy_srq(ndev);
+ kfree(ndev);
+}
+
+static int nvmet_ib_pool_dev_init(struct ib_pool_device *dev)
+{
+ struct nvmet_rdma_device *ndev;
+
+ ndev = container_of(dev, typeof(*ndev), dev);
+ if (nvmet_rdma_use_srq)
+ return nvmet_rdma_init_srq(ndev);
+
+ return 0;
+}
+
static int __init nvmet_rdma_init(void)
{
int ret;
+ ib_pool_dev_init(0, nvmet_ib_pool_dev_alloc, nvmet_ib_pool_dev_free,
+ nvmet_ib_pool_dev_init, &ib_devs_pool);
+
ret = ib_register_client(&nvmet_rdma_ib_client);
if (ret)
return ret;
@@ -1533,6 +1496,7 @@ static void __exit nvmet_rdma_exit(void)
ib_unregister_client(&nvmet_rdma_ib_client);
WARN_ON_ONCE(!list_empty(&nvmet_rdma_queue_list));
ida_destroy(&nvmet_rdma_queue_ida);
+ ib_pool_dev_deinit(&ib_devs_pool);
}
module_init(nvmet_rdma_init);
--
2.13.1
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
