Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1285922 > unrolled thread

completion queue abstraction V2

Started byChristoph Hellwig <hch@lst.de>
First post2015-12-07 22:00 +0100
Last post2015-12-07 22:10 +0100
Articles 7 — 1 participant

Back to article view | Back to linux.kernel

Contents

  completion queue abstraction V2 Christoph Hellwig <hch@lst.de> - 2015-12-07 22:00 +0100
    [PATCH 06/13] irq_poll: remove unused data and max fields Christoph Hellwig <hch@lst.de> - 2015-12-07 22:00 +0100
    [PATCH 09/13] IB/srpt: use the new CQ API Christoph Hellwig <hch@lst.de> - 2015-12-07 22:00 +0100
    [PATCH 03/13] irq_poll: fold irq_poll_sched_prep into irq_poll_sched Christoph Hellwig <hch@lst.de> - 2015-12-07 22:10 +0100
    [PATCH 04/13] irq_poll: fold irq_poll_disable_pending into irq_poll_softirq Christoph Hellwig <hch@lst.de> - 2015-12-07 22:10 +0100
    [PATCH 02/13] irq_poll: don't disable new irq_poll instances Christoph Hellwig <hch@lst.de> - 2015-12-07 22:10 +0100
    [PATCH 01/13] irq_poll: make blk-iopoll available outside the block layer Christoph Hellwig <hch@lst.de> - 2015-12-07 22:10 +0100

#1285922 — completion queue abstraction V2

FromChristoph Hellwig <hch@lst.de>
Date2015-12-07 22:00 +0100
Subjectcompletion queue abstraction V2
Message-ID<qDb34-14n-3@gated-at.bofh.it>
This series adds a new RDMA core abstraction that insulated the
ULPs from the nitty gritty details of CQ polling.  See the individual
patches for more details.

Note that this series should be applied on top of my
"IB: merge struct ib_device_attr into struct ib_device" patch and the
MR cleanups.

A git tree is also available:

	git://git.infradead.org/users/hch/rdma.git rdma-cq

As well as gitweb:
	http://git.infradead.org/users/hch/rdma.git/shortlog/refs/heads/rdma-cq

Changes since V1:
 - rebased
 - some additional irq_work cleanups
 - early exit from the polling loop (Jason)
 - changed ib_process_cq_direct API (Sagi / Jason / Bart)
 - dropped the ib_drain_qp helper for now, to be revisited later
 - cosmetic iser cleanups (or)
 - cosmetic srp changes (Bart)
 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [next] | [standalone]

#1285924 — [PATCH 06/13] irq_poll: remove unused data and max fields

FromChristoph Hellwig <hch@lst.de>
Date2015-12-07 22:00 +0100
Subject[PATCH 06/13] irq_poll: remove unused data and max fields
Message-ID<qDb35-14n-35@gated-at.bofh.it>
In reply to#1285922
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/irq_poll.h | 2 --
 1 file changed, 2 deletions(-)

diff --git a/include/linux/irq_poll.h b/include/linux/irq_poll.h
index ce9e1db..7527c03 100644
--- a/include/linux/irq_poll.h
+++ b/include/linux/irq_poll.h
@@ -7,9 +7,7 @@ typedef int (irq_poll_fn)(struct irq_poll *, int);
 struct irq_poll {
 	struct list_head list;
 	unsigned long state;
-	unsigned long data;
 	int weight;
-	int max;
 	irq_poll_fn *poll;
 };
 
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1285925 — [PATCH 09/13] IB/srpt: use the new CQ API

FromChristoph Hellwig <hch@lst.de>
Date2015-12-07 22:00 +0100
Subject[PATCH 09/13] IB/srpt: use the new CQ API
Message-ID<qDb35-14n-37@gated-at.bofh.it>
In reply to#1285922
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/infiniband/ulp/srpt/ib_srpt.c | 327 +++++++++-------------------------
 drivers/infiniband/ulp/srpt/ib_srpt.h |  28 +--
 2 files changed, 88 insertions(+), 267 deletions(-)

diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c
index b5544b2..98282e9 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.c
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.c
@@ -93,6 +93,8 @@ MODULE_PARM_DESC(srpt_service_guid,
 static struct ib_client srpt_client;
 static void srpt_release_channel(struct srpt_rdma_ch *ch);
 static int srpt_queue_status(struct se_cmd *cmd);
+static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc);
+static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc);
 
 /**
  * opposite_dma_dir() - Swap DMA_TO_DEVICE and DMA_FROM_DEVICE.
@@ -778,12 +780,12 @@ static int srpt_post_recv(struct srpt_device *sdev,
 	struct ib_recv_wr wr, *bad_wr;
 
 	BUG_ON(!sdev);
-	wr.wr_id = encode_wr_id(SRPT_RECV, ioctx->ioctx.index);
-
 	list.addr = ioctx->ioctx.dma;
 	list.length = srp_max_req_size;
 	list.lkey = sdev->pd->local_dma_lkey;
 
+	ioctx->ioctx.cqe.done = srpt_recv_done;
+	wr.wr_cqe = &ioctx->ioctx.cqe;
 	wr.next = NULL;
 	wr.sg_list = &list;
 	wr.num_sge = 1;
@@ -819,8 +821,9 @@ static int srpt_post_send(struct srpt_rdma_ch *ch,
 	list.length = len;
 	list.lkey = sdev->pd->local_dma_lkey;
 
+	ioctx->ioctx.cqe.done = srpt_send_done;
 	wr.next = NULL;
-	wr.wr_id = encode_wr_id(SRPT_SEND, ioctx->ioctx.index);
+	wr.wr_cqe = &ioctx->ioctx.cqe;
 	wr.sg_list = &list;
 	wr.num_sge = 1;
 	wr.opcode = IB_WR_SEND;
@@ -1383,116 +1386,44 @@ out:
 }
 
 /**
- * srpt_handle_send_err_comp() - Process an IB_WC_SEND error completion.
- */
-static void srpt_handle_send_err_comp(struct srpt_rdma_ch *ch, u64 wr_id)
-{
-	struct srpt_send_ioctx *ioctx;
-	enum srpt_command_state state;
-	u32 index;
-
-	atomic_inc(&ch->sq_wr_avail);
-
-	index = idx_from_wr_id(wr_id);
-	ioctx = ch->ioctx_ring[index];
-	state = srpt_get_cmd_state(ioctx);
-
-	WARN_ON(state != SRPT_STATE_CMD_RSP_SENT
-		&& state != SRPT_STATE_MGMT_RSP_SENT
-		&& state != SRPT_STATE_NEED_DATA
-		&& state != SRPT_STATE_DONE);
-
-	/* If SRP_RSP sending failed, undo the ch->req_lim change. */
-	if (state == SRPT_STATE_CMD_RSP_SENT
-	    || state == SRPT_STATE_MGMT_RSP_SENT)
-		atomic_dec(&ch->req_lim);
-
-	srpt_abort_cmd(ioctx);
-}
-
-/**
- * srpt_handle_send_comp() - Process an IB send completion notification.
- */
-static void srpt_handle_send_comp(struct srpt_rdma_ch *ch,
-				  struct srpt_send_ioctx *ioctx)
-{
-	enum srpt_command_state state;
-
-	atomic_inc(&ch->sq_wr_avail);
-
-	state = srpt_set_cmd_state(ioctx, SRPT_STATE_DONE);
-
-	if (WARN_ON(state != SRPT_STATE_CMD_RSP_SENT
-		    && state != SRPT_STATE_MGMT_RSP_SENT
-		    && state != SRPT_STATE_DONE))
-		pr_debug("state = %d\n", state);
-
-	if (state != SRPT_STATE_DONE) {
-		srpt_unmap_sg_to_ib_sge(ch, ioctx);
-		transport_generic_free_cmd(&ioctx->cmd, 0);
-	} else {
-		pr_err("IB completion has been received too late for"
-		       " wr_id = %u.\n", ioctx->ioctx.index);
-	}
-}
-
-/**
- * srpt_handle_rdma_comp() - Process an IB RDMA completion notification.
- *
  * XXX: what is now target_execute_cmd used to be asynchronous, and unmapping
  * the data that has been transferred via IB RDMA had to be postponed until the
  * check_stop_free() callback.  None of this is necessary anymore and needs to
  * be cleaned up.
  */
-static void srpt_handle_rdma_comp(struct srpt_rdma_ch *ch,
-				  struct srpt_send_ioctx *ioctx,
-				  enum srpt_opcode opcode)
+static void srpt_rdma_read_done(struct ib_cq *cq, struct ib_wc *wc)
 {
+	struct srpt_rdma_ch *ch = cq->cq_context;
+	struct srpt_send_ioctx *ioctx =
+		container_of(wc->wr_cqe, struct srpt_send_ioctx, ioctx.cqe);
+
 	WARN_ON(ioctx->n_rdma <= 0);
 	atomic_add(ioctx->n_rdma, &ch->sq_wr_avail);
 
-	if (opcode == SRPT_RDMA_READ_LAST) {
-		if (srpt_test_and_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA,
-						SRPT_STATE_DATA_IN))
-			target_execute_cmd(&ioctx->cmd);
-		else
-			pr_err("%s[%d]: wrong state = %d\n", __func__,
-			       __LINE__, srpt_get_cmd_state(ioctx));
-	} else {
-		WARN(true, "unexpected opcode %d\n", opcode);
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		pr_info("RDMA_READ for ioctx 0x%p failed with status %d\n",
+			ioctx, wc->status);
+		srpt_abort_cmd(ioctx);
+		return;
 	}
+
+	if (srpt_test_and_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA,
+					SRPT_STATE_DATA_IN))
+		target_execute_cmd(&ioctx->cmd);
+	else
+		pr_err("%s[%d]: wrong state = %d\n", __func__,
+		       __LINE__, srpt_get_cmd_state(ioctx));
 }
 
-/**
- * srpt_handle_rdma_err_comp() - Process an IB RDMA error completion.
- */
-static void srpt_handle_rdma_err_comp(struct srpt_rdma_ch *ch,
-				      struct srpt_send_ioctx *ioctx,
-				      enum srpt_opcode opcode)
+static void srpt_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc)
 {
-	enum srpt_command_state state;
+	struct srpt_send_ioctx *ioctx =
+		container_of(wc->wr_cqe, struct srpt_send_ioctx, ioctx.cqe);
 
-	state = srpt_get_cmd_state(ioctx);
-	switch (opcode) {
-	case SRPT_RDMA_READ_LAST:
-		if (ioctx->n_rdma <= 0) {
-			pr_err("Received invalid RDMA read"
-			       " error completion with idx %d\n",
-			       ioctx->ioctx.index);
-			break;
-		}
-		atomic_add(ioctx->n_rdma, &ch->sq_wr_avail);
-		if (state == SRPT_STATE_NEED_DATA)
-			srpt_abort_cmd(ioctx);
-		else
-			pr_err("%s[%d]: wrong state = %d\n",
-			       __func__, __LINE__, state);
-		break;
-	case SRPT_RDMA_WRITE_LAST:
-		break;
-	default:
-		pr_err("%s[%d]: opcode = %u\n", __func__, __LINE__, opcode);
-		break;
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		pr_info("RDMA_WRITE for ioctx 0x%p failed with status %d\n",
+			ioctx, wc->status);
+		srpt_abort_cmd(ioctx);
 	}
 }
 
@@ -1927,32 +1858,26 @@ out:
 	return;
 }
 
-static void srpt_process_rcv_completion(struct ib_cq *cq,
-					struct srpt_rdma_ch *ch,
-					struct ib_wc *wc)
+static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct srpt_device *sdev = ch->sport->sdev;
-	struct srpt_recv_ioctx *ioctx;
-	u32 index;
+	struct srpt_rdma_ch *ch = cq->cq_context;
+	struct srpt_recv_ioctx *ioctx =
+		container_of(wc->wr_cqe, struct srpt_recv_ioctx, ioctx.cqe);
 
-	index = idx_from_wr_id(wc->wr_id);
 	if (wc->status == IB_WC_SUCCESS) {
 		int req_lim;
 
 		req_lim = atomic_dec_return(&ch->req_lim);
 		if (unlikely(req_lim < 0))
 			pr_err("req_lim = %d < 0\n", req_lim);
-		ioctx = sdev->ioctx_ring[index];
 		srpt_handle_new_iu(ch, ioctx, NULL);
 	} else {
-		pr_info("receiving failed for idx %u with status %d\n",
-			index, wc->status);
+		pr_info("receiving failed for ioctx %p with status %d\n",
+			ioctx, wc->status);
 	}
 }
 
 /**
- * srpt_process_send_completion() - Process an IB send completion.
- *
  * Note: Although this has not yet been observed during tests, at least in
  * theory it is possible that the srpt_get_send_ioctx() call invoked by
  * srpt_handle_new_iu() fails. This is possible because the req_lim_delta
@@ -1965,108 +1890,52 @@ static void srpt_process_rcv_completion(struct ib_cq *cq,
  * are queued on cmd_wait_list. The code below processes these delayed
  * requests one at a time.
  */
-static void srpt_process_send_completion(struct ib_cq *cq,
-					 struct srpt_rdma_ch *ch,
-					 struct ib_wc *wc)
+static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct srpt_send_ioctx *send_ioctx;
-	uint32_t index;
-	enum srpt_opcode opcode;
+	struct srpt_rdma_ch *ch = cq->cq_context;
+	struct srpt_send_ioctx *ioctx =
+		container_of(wc->wr_cqe, struct srpt_send_ioctx, ioctx.cqe);
+	enum srpt_command_state state;
 
-	index = idx_from_wr_id(wc->wr_id);
-	opcode = opcode_from_wr_id(wc->wr_id);
-	send_ioctx = ch->ioctx_ring[index];
-	if (wc->status == IB_WC_SUCCESS) {
-		if (opcode == SRPT_SEND)
-			srpt_handle_send_comp(ch, send_ioctx);
-		else {
-			WARN_ON(wc->opcode != IB_WC_RDMA_READ);
-			srpt_handle_rdma_comp(ch, send_ioctx, opcode);
-		}
+	state = srpt_set_cmd_state(ioctx, SRPT_STATE_DONE);
+
+	WARN_ON(state != SRPT_STATE_CMD_RSP_SENT &&
+		state != SRPT_STATE_MGMT_RSP_SENT);
+
+	atomic_inc(&ch->sq_wr_avail);
+
+	if (wc->status != IB_WC_SUCCESS) {
+		pr_info("sending response for ioctx 0x%p failed"
+			" with status %d\n", ioctx, wc->status);
+
+		atomic_dec(&ch->req_lim);
+		srpt_abort_cmd(ioctx);
+		goto out;
+	}
+
+	if (state != SRPT_STATE_DONE) {
+		srpt_unmap_sg_to_ib_sge(ch, ioctx);
+		transport_generic_free_cmd(&ioctx->cmd, 0);
 	} else {
-		if (opcode == SRPT_SEND) {
-			pr_info("sending response for idx %u failed"
-				" with status %d\n", index, wc->status);
-			srpt_handle_send_err_comp(ch, wc->wr_id);
-		} else if (opcode != SRPT_RDMA_MID) {
-			pr_info("RDMA t %d for idx %u failed with"
-				" status %d\n", opcode, index, wc->status);
-			srpt_handle_rdma_err_comp(ch, send_ioctx, opcode);
-		}
+		pr_err("IB completion has been received too late for"
+		       " wr_id = %u.\n", ioctx->ioctx.index);
 	}
 
-	while (unlikely(opcode == SRPT_SEND
-			&& !list_empty(&ch->cmd_wait_list)
-			&& srpt_get_ch_state(ch) == CH_LIVE
-			&& (send_ioctx = srpt_get_send_ioctx(ch)) != NULL)) {
+out:
+	while (!list_empty(&ch->cmd_wait_list) &&
+	       srpt_get_ch_state(ch) == CH_LIVE &&
+	       (ioctx = srpt_get_send_ioctx(ch)) != NULL) {
 		struct srpt_recv_ioctx *recv_ioctx;
 
 		recv_ioctx = list_first_entry(&ch->cmd_wait_list,
 					      struct srpt_recv_ioctx,
 					      wait_list);
 		list_del(&recv_ioctx->wait_list);
-		srpt_handle_new_iu(ch, recv_ioctx, send_ioctx);
-	}
-}
-
-static void srpt_process_completion(struct ib_cq *cq, struct srpt_rdma_ch *ch)
-{
-	struct ib_wc *const wc = ch->wc;
-	int i, n;
-
-	WARN_ON(cq != ch->cq);
-
-	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
-	while ((n = ib_poll_cq(cq, ARRAY_SIZE(ch->wc), wc)) > 0) {
-		for (i = 0; i < n; i++) {
-			if (opcode_from_wr_id(wc[i].wr_id) == SRPT_RECV)
-				srpt_process_rcv_completion(cq, ch, &wc[i]);
-			else
-				srpt_process_send_completion(cq, ch, &wc[i]);
-		}
+		srpt_handle_new_iu(ch, recv_ioctx, ioctx);
 	}
 }
 
 /**
- * srpt_completion() - IB completion queue callback function.
- *
- * Notes:
- * - It is guaranteed that a completion handler will never be invoked
- *   concurrently on two different CPUs for the same completion queue. See also
- *   Documentation/infiniband/core_locking.txt and the implementation of
- *   handle_edge_irq() in kernel/irq/chip.c.
- * - When threaded IRQs are enabled, completion handlers are invoked in thread
- *   context instead of interrupt context.
- */
-static void srpt_completion(struct ib_cq *cq, void *ctx)
-{
-	struct srpt_rdma_ch *ch = ctx;
-
-	wake_up_interruptible(&ch->wait_queue);
-}
-
-static int srpt_compl_thread(void *arg)
-{
-	struct srpt_rdma_ch *ch;
-
-	/* Hibernation / freezing of the SRPT kernel thread is not supported. */
-	current->flags |= PF_NOFREEZE;
-
-	ch = arg;
-	BUG_ON(!ch);
-	pr_info("Session %s: kernel thread %s (PID %d) started\n",
-		ch->sess_name, ch->thread->comm, current->pid);
-	while (!kthread_should_stop()) {
-		wait_event_interruptible(ch->wait_queue,
-			(srpt_process_completion(ch->cq, ch),
-			 kthread_should_stop()));
-	}
-	pr_info("Session %s: kernel thread %s (PID %d) stopped\n",
-		ch->sess_name, ch->thread->comm, current->pid);
-	return 0;
-}
-
-/**
  * srpt_create_ch_ib() - Create receive and send completion queues.
  */
 static int srpt_create_ch_ib(struct srpt_rdma_ch *ch)
@@ -2075,7 +1944,6 @@ static int srpt_create_ch_ib(struct srpt_rdma_ch *ch)
 	struct srpt_port *sport = ch->sport;
 	struct srpt_device *sdev = sport->sdev;
 	u32 srp_sq_size = sport->port_attrib.srp_sq_size;
-	struct ib_cq_init_attr cq_attr = {};
 	int ret;
 
 	WARN_ON(ch->rq_size < 1);
@@ -2086,9 +1954,8 @@ static int srpt_create_ch_ib(struct srpt_rdma_ch *ch)
 		goto out;
 
 retry:
-	cq_attr.cqe = ch->rq_size + srp_sq_size;
-	ch->cq = ib_create_cq(sdev->device, srpt_completion, NULL, ch,
-			      &cq_attr);
+	ch->cq = ib_alloc_cq(sdev->device, ch, ch->rq_size + srp_sq_size,
+			0 /* XXX: spread CQs */, IB_POLL_WORKQUEUE);
 	if (IS_ERR(ch->cq)) {
 		ret = PTR_ERR(ch->cq);
 		pr_err("failed to create CQ cqe= %d ret= %d\n",
@@ -2131,18 +1998,6 @@ retry:
 	if (ret)
 		goto err_destroy_qp;
 
-	init_waitqueue_head(&ch->wait_queue);
-
-	pr_debug("creating thread for session %s\n", ch->sess_name);
-
-	ch->thread = kthread_run(srpt_compl_thread, ch, "ib_srpt_compl");
-	if (IS_ERR(ch->thread)) {
-		pr_err("failed to create kernel thread %ld\n",
-		       PTR_ERR(ch->thread));
-		ch->thread = NULL;
-		goto err_destroy_qp;
-	}
-
 out:
 	kfree(qp_init);
 	return ret;
@@ -2150,17 +2005,14 @@ out:
 err_destroy_qp:
 	ib_destroy_qp(ch->qp);
 err_destroy_cq:
-	ib_destroy_cq(ch->cq);
+	ib_free_cq(ch->cq);
 	goto out;
 }
 
 static void srpt_destroy_ch_ib(struct srpt_rdma_ch *ch)
 {
-	if (ch->thread)
-		kthread_stop(ch->thread);
-
 	ib_destroy_qp(ch->qp);
-	ib_destroy_cq(ch->cq);
+	ib_free_cq(ch->cq);
 }
 
 /**
@@ -2822,9 +2674,7 @@ static int srpt_perform_rdmas(struct srpt_rdma_ch *ch,
 			      struct srpt_send_ioctx *ioctx)
 {
 	struct ib_send_wr *bad_wr;
-	int i;
-	int ret;
-	int sq_wr_avail;
+	int sq_wr_avail, ret, i;
 	enum dma_data_direction dir;
 	const int n_rdma = ioctx->n_rdma;
 
@@ -2841,29 +2691,24 @@ static int srpt_perform_rdmas(struct srpt_rdma_ch *ch,
 	}
 
 	for (i = 0; i < n_rdma; i++) {
-		struct ib_rdma_wr *wr = &ioctx->rdma_ius[i];
-
-		if (dir == DMA_FROM_DEVICE) {
-			wr->wr.opcode = IB_WR_RDMA_WRITE;
-			wr->wr.wr_id = encode_wr_id(i == n_rdma - 1 ?
-						SRPT_RDMA_WRITE_LAST :
-						SRPT_RDMA_MID,
-						ioctx->ioctx.index);
-		} else {
-			wr->wr.opcode = IB_WR_RDMA_READ;
-			wr->wr.wr_id = encode_wr_id(i == n_rdma - 1 ?
-						SRPT_RDMA_READ_LAST :
-						SRPT_RDMA_MID,
-						ioctx->ioctx.index);
-		}
+		struct ib_send_wr *wr = &ioctx->rdma_ius[i].wr;
+
+		wr->opcode = (dir == DMA_FROM_DEVICE) ?
+				IB_WR_RDMA_WRITE : IB_WR_RDMA_READ;
 
 		if (i == n_rdma - 1) {
 			/* only get completion event for the last rdma read */
-			if (dir == DMA_TO_DEVICE)
-				wr->wr.send_flags = IB_SEND_SIGNALED;
-			wr->wr.next = NULL;
+			if (dir == DMA_TO_DEVICE) {
+				wr->send_flags = IB_SEND_SIGNALED;
+				ioctx->rdma_cqe.done = srpt_rdma_read_done;
+			} else {
+				ioctx->rdma_cqe.done = srpt_rdma_write_done;
+			}
+			wr->wr_cqe = &ioctx->rdma_cqe;
+			wr->next = NULL;
 		} else {
-			wr->wr.next = &ioctx->rdma_ius[i + 1].wr;
+			wr->wr_cqe = NULL;
+			wr->next = &ioctx->rdma_ius[i + 1].wr;
 		}
 	}
 
diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.h b/drivers/infiniband/ulp/srpt/ib_srpt.h
index fd6097e..f9568f5 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.h
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.h
@@ -128,27 +128,6 @@ enum {
 	DEFAULT_MAX_RDMA_SIZE = 65536,
 };
 
-enum srpt_opcode {
-	SRPT_RECV,
-	SRPT_SEND,
-	SRPT_RDMA_MID,
-	SRPT_RDMA_READ_LAST,
-	SRPT_RDMA_WRITE_LAST,
-};
-
-static inline u64 encode_wr_id(u8 opcode, u32 idx)
-{
-	return ((u64)opcode << 32) | idx;
-}
-static inline enum srpt_opcode opcode_from_wr_id(u64 wr_id)
-{
-	return wr_id >> 32;
-}
-static inline u32 idx_from_wr_id(u64 wr_id)
-{
-	return (u32)wr_id;
-}
-
 /**
  * enum srpt_command_state - SCSI command state managed by SRPT.
  * @SRPT_STATE_NEW:           New command arrived and is being processed.
@@ -180,6 +159,7 @@ enum srpt_command_state {
  * @index: Index of the I/O context in its ioctx_ring array.
  */
 struct srpt_ioctx {
+	struct ib_cqe		cqe;
 	void			*buf;
 	dma_addr_t		dma;
 	uint32_t		index;
@@ -218,6 +198,7 @@ struct srpt_send_ioctx {
 	struct srpt_ioctx	ioctx;
 	struct srpt_rdma_ch	*ch;
 	struct ib_rdma_wr	*rdma_ius;
+	struct ib_cqe		rdma_cqe;
 	struct srp_direct_buf	*rbufs;
 	struct srp_direct_buf	single_rbuf;
 	struct scatterlist	*sg;
@@ -255,9 +236,6 @@ enum rdma_ch_state {
 
 /**
  * struct srpt_rdma_ch - RDMA channel.
- * @wait_queue:    Allows the kernel thread to wait for more work.
- * @thread:        Kernel thread that processes the IB queues associated with
- *                 the channel.
  * @cm_id:         IB CM ID associated with the channel.
  * @qp:            IB queue pair used for communicating over this channel.
  * @cq:            IB completion queue for this channel.
@@ -287,8 +265,6 @@ enum rdma_ch_state {
  * @release_done:  Enables waiting for srpt_release_channel() completion.
  */
 struct srpt_rdma_ch {
-	wait_queue_head_t	wait_queue;
-	struct task_struct	*thread;
 	struct ib_cm_id		*cm_id;
 	struct ib_qp		*qp;
 	struct ib_cq		*cq;
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1285926 — [PATCH 03/13] irq_poll: fold irq_poll_sched_prep into irq_poll_sched

FromChristoph Hellwig <hch@lst.de>
Date2015-12-07 22:10 +0100
Subject[PATCH 03/13] irq_poll: fold irq_poll_sched_prep into irq_poll_sched
Message-ID<qDbcJ-1od-1@gated-at.bofh.it>
In reply to#1285922
There is no good reason to keep them apart, and this makes using the API
a bit simpler.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/be2iscsi/be_main.c |  6 ++----
 drivers/scsi/ipr.c              |  3 +--
 include/linux/irq_poll.h        | 13 -------------
 lib/irq_poll.c                  | 10 +++++++---
 4 files changed, 10 insertions(+), 22 deletions(-)

diff --git a/drivers/scsi/be2iscsi/be_main.c b/drivers/scsi/be2iscsi/be_main.c
index 471e2b9..cb9072a8 100644
--- a/drivers/scsi/be2iscsi/be_main.c
+++ b/drivers/scsi/be2iscsi/be_main.c
@@ -910,8 +910,7 @@ static irqreturn_t be_isr_msix(int irq, void *dev_id)
 	num_eq_processed = 0;
 	while (eqe->dw[offsetof(struct amap_eq_entry, valid) / 32]
 				& EQE_VALID_MASK) {
-		if (!irq_poll_sched_prep(&pbe_eq->iopoll))
-			irq_poll_sched(&pbe_eq->iopoll);
+		irq_poll_sched(&pbe_eq->iopoll);
 
 		AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
 		queue_tail_inc(eq);
@@ -972,8 +971,7 @@ static irqreturn_t be_isr(int irq, void *dev_id)
 			spin_unlock_irqrestore(&phba->isr_lock, flags);
 			num_mcceq_processed++;
 		} else {
-			if (!irq_poll_sched_prep(&pbe_eq->iopoll))
-				irq_poll_sched(&pbe_eq->iopoll);
+			irq_poll_sched(&pbe_eq->iopoll);
 			num_ioeq_processed++;
 		}
 		AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
diff --git a/drivers/scsi/ipr.c b/drivers/scsi/ipr.c
index 402e4ca..82031e0 100644
--- a/drivers/scsi/ipr.c
+++ b/drivers/scsi/ipr.c
@@ -5692,8 +5692,7 @@ static irqreturn_t ipr_isr_mhrrq(int irq, void *devp)
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		if ((be32_to_cpu(*hrrq->hrrq_curr) & IPR_HRRQ_TOGGLE_BIT) ==
 		       hrrq->toggle_bit) {
-			if (!irq_poll_sched_prep(&hrrq->iopoll))
-				irq_poll_sched(&hrrq->iopoll);
+			irq_poll_sched(&hrrq->iopoll);
 			spin_unlock_irqrestore(hrrq->lock, hrrq_flags);
 			return IRQ_HANDLED;
 		}
diff --git a/include/linux/irq_poll.h b/include/linux/irq_poll.h
index 0cf7c26..73d7c20 100644
--- a/include/linux/irq_poll.h
+++ b/include/linux/irq_poll.h
@@ -18,19 +18,6 @@ enum {
 	IRQ_POLL_F_DISABLE	= 1,
 };
 
-/*
- * Returns 0 if we successfully set the IRQ_POLL_F_SCHED bit, indicating
- * that we were the first to acquire this iop for scheduling. If this iop
- * is currently disabled, return "failure".
- */
-static inline int irq_poll_sched_prep(struct irq_poll *iop)
-{
-	if (!test_bit(IRQ_POLL_F_DISABLE, &iop->state))
-		return test_and_set_bit(IRQ_POLL_F_SCHED, &iop->state);
-
-	return 1;
-}
-
 static inline int irq_poll_disable_pending(struct irq_poll *iop)
 {
 	return test_bit(IRQ_POLL_F_DISABLE, &iop->state);
diff --git a/lib/irq_poll.c b/lib/irq_poll.c
index 88af879..13cb149 100644
--- a/lib/irq_poll.c
+++ b/lib/irq_poll.c
@@ -21,13 +21,17 @@ static DEFINE_PER_CPU(struct list_head, blk_cpu_iopoll);
  *
  * Description:
  *     Add this irq_poll structure to the pending poll list and trigger the
- *     raise of the blk iopoll softirq. The driver must already have gotten a
- *     successful return from irq_poll_sched_prep() before calling this.
+ *     raise of the blk iopoll softirq.
  **/
 void irq_poll_sched(struct irq_poll *iop)
 {
 	unsigned long flags;
 
+	if (test_bit(IRQ_POLL_F_DISABLE, &iop->state))
+		return;
+	if (!test_and_set_bit(IRQ_POLL_F_SCHED, &iop->state))
+		return;
+
 	local_irq_save(flags);
 	list_add_tail(&iop->list, this_cpu_ptr(&blk_cpu_iopoll));
 	__raise_softirq_irqoff(IRQ_POLL_SOFTIRQ);
@@ -58,7 +62,7 @@ EXPORT_SYMBOL(__irq_poll_complete);
  * Description:
  *     If a driver consumes less than the assigned budget in its run of the
  *     iopoll handler, it'll end the polled mode by calling this function. The
- *     iopoll handler will not be invoked again before irq_poll_sched_prep()
+ *     iopoll handler will not be invoked again before irq_poll_schedp()
  *     is called.
  **/
 void irq_poll_complete(struct irq_poll *iop)
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1285927 — [PATCH 04/13] irq_poll: fold irq_poll_disable_pending into irq_poll_softirq

FromChristoph Hellwig <hch@lst.de>
Date2015-12-07 22:10 +0100
Subject[PATCH 04/13] irq_poll: fold irq_poll_disable_pending into irq_poll_softirq
Message-ID<qDbcJ-1od-3@gated-at.bofh.it>
In reply to#1285922
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/irq_poll.h | 5 -----
 lib/irq_poll.c           | 2 +-
 2 files changed, 1 insertion(+), 6 deletions(-)

diff --git a/include/linux/irq_poll.h b/include/linux/irq_poll.h
index 73d7c20..c3145c7 100644
--- a/include/linux/irq_poll.h
+++ b/include/linux/irq_poll.h
@@ -18,11 +18,6 @@ enum {
 	IRQ_POLL_F_DISABLE	= 1,
 };
 
-static inline int irq_poll_disable_pending(struct irq_poll *iop)
-{
-	return test_bit(IRQ_POLL_F_DISABLE, &iop->state);
-}
-
 extern void irq_poll_sched(struct irq_poll *);
 extern void irq_poll_init(struct irq_poll *, int, irq_poll_fn *);
 extern void irq_poll_complete(struct irq_poll *);
diff --git a/lib/irq_poll.c b/lib/irq_poll.c
index 13cb149..2c7b8e6 100644
--- a/lib/irq_poll.c
+++ b/lib/irq_poll.c
@@ -122,7 +122,7 @@ static void irq_poll_softirq(struct softirq_action *h)
 		 * move the instance around on the list at-will.
 		 */
 		if (work >= weight) {
-			if (irq_poll_disable_pending(iop))
+			if (test_bit(IRQ_POLL_F_DISABLE, &iop->state))
 				__irq_poll_complete(iop);
 			else
 				list_move_tail(&iop->list, list);
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1285930 — [PATCH 02/13] irq_poll: don't disable new irq_poll instances

FromChristoph Hellwig <hch@lst.de>
Date2015-12-07 22:10 +0100
Subject[PATCH 02/13] irq_poll: don't disable new irq_poll instances
Message-ID<qDbcK-1od-13@gated-at.bofh.it>
In reply to#1285922
There is no good reason to start out disabled - drivers can control if
the poll instance can be scheduled by simply not scheduling it yet.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/be2iscsi/be_main.c | 2 --
 drivers/scsi/ipr.c              | 2 --
 lib/irq_poll.c                  | 4 +---
 3 files changed, 1 insertion(+), 7 deletions(-)

diff --git a/drivers/scsi/be2iscsi/be_main.c b/drivers/scsi/be2iscsi/be_main.c
index 1d879ef..471e2b9 100644
--- a/drivers/scsi/be2iscsi/be_main.c
+++ b/drivers/scsi/be2iscsi/be_main.c
@@ -5581,7 +5581,6 @@ static void beiscsi_eeh_resume(struct pci_dev *pdev)
 		pbe_eq = &phwi_context->be_eq[i];
 		irq_poll_init(&pbe_eq->iopoll, be_iopoll_budget,
 				be_iopoll);
-		irq_poll_enable(&pbe_eq->iopoll);
 	}
 
 	i = (phba->msix_enabled) ? i : 0;
@@ -5754,7 +5753,6 @@ static int beiscsi_dev_probe(struct pci_dev *pcidev,
 		pbe_eq = &phwi_context->be_eq[i];
 		irq_poll_init(&pbe_eq->iopoll, be_iopoll_budget,
 				be_iopoll);
-		irq_poll_enable(&pbe_eq->iopoll);
 	}
 
 	i = (phba->msix_enabled) ? i : 0;
diff --git a/drivers/scsi/ipr.c b/drivers/scsi/ipr.c
index 6b9c738..402e4ca 100644
--- a/drivers/scsi/ipr.c
+++ b/drivers/scsi/ipr.c
@@ -3708,7 +3708,6 @@ static ssize_t ipr_store_iopoll_weight(struct device *dev,
 		for (i = 1; i < ioa_cfg->hrrq_num; i++) {
 			irq_poll_init(&ioa_cfg->hrrq[i].iopoll,
 					ioa_cfg->iopoll_weight, ipr_iopoll);
-			irq_poll_enable(&ioa_cfg->hrrq[i].iopoll);
 		}
 	}
 	spin_unlock_irqrestore(shost->host_lock, lock_flags);
@@ -10407,7 +10406,6 @@ static int ipr_probe(struct pci_dev *pdev, const struct pci_device_id *dev_id)
 		for (i = 1; i < ioa_cfg->hrrq_num; i++) {
 			irq_poll_init(&ioa_cfg->hrrq[i].iopoll,
 					ioa_cfg->iopoll_weight, ipr_iopoll);
-			irq_poll_enable(&ioa_cfg->hrrq[i].iopoll);
 		}
 	}
 
diff --git a/lib/irq_poll.c b/lib/irq_poll.c
index e6fd1dc..88af879 100644
--- a/lib/irq_poll.c
+++ b/lib/irq_poll.c
@@ -170,8 +170,7 @@ EXPORT_SYMBOL(irq_poll_enable);
  * @poll_fn:  The handler to invoke
  *
  * Description:
- *     Initialize this irq_poll structure. Before being actively used, the
- *     driver must call irq_poll_enable().
+ *     Initialize and enable this irq_poll structure.
  **/
 void irq_poll_init(struct irq_poll *iop, int weight, irq_poll_fn *poll_fn)
 {
@@ -179,7 +178,6 @@ void irq_poll_init(struct irq_poll *iop, int weight, irq_poll_fn *poll_fn)
 	INIT_LIST_HEAD(&iop->list);
 	iop->weight = weight;
 	iop->poll = poll_fn;
-	set_bit(IRQ_POLL_F_SCHED, &iop->state);
 }
 EXPORT_SYMBOL(irq_poll_init);
 
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1285931 — [PATCH 01/13] irq_poll: make blk-iopoll available outside the block layer

FromChristoph Hellwig <hch@lst.de>
Date2015-12-07 22:10 +0100
Subject[PATCH 01/13] irq_poll: make blk-iopoll available outside the block layer
Message-ID<qDbcK-1od-17@gated-at.bofh.it>
In reply to#1285922
The new name is irq_poll as iopoll is already taken.  Better suggestions
welcome.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 Documentation/kernel-per-CPU-kthreads.txt |   2 +-
 block/Makefile                            |   2 +-
 block/blk-iopoll.c                        | 224 ------------------------------
 drivers/scsi/Kconfig                      |   1 +
 drivers/scsi/be2iscsi/Kconfig             |   1 +
 drivers/scsi/be2iscsi/be.h                |   4 +-
 drivers/scsi/be2iscsi/be_iscsi.c          |   4 +-
 drivers/scsi/be2iscsi/be_main.c           |  24 ++--
 drivers/scsi/ipr.c                        |  28 ++--
 drivers/scsi/ipr.h                        |   4 +-
 include/linux/blk-iopoll.h                |  46 ------
 include/linux/interrupt.h                 |   2 +-
 include/linux/irq_poll.h                  |  46 ++++++
 include/trace/events/irq.h                |   2 +-
 lib/Kconfig                               |   5 +
 lib/Makefile                              |   1 +
 lib/irq_poll.c                            | 221 +++++++++++++++++++++++++++++
 tools/lib/traceevent/event-parse.c        |   2 +-
 tools/perf/util/trace-event-parse.c       |   2 +-
 19 files changed, 313 insertions(+), 308 deletions(-)
 delete mode 100644 block/blk-iopoll.c
 delete mode 100644 include/linux/blk-iopoll.h
 create mode 100644 include/linux/irq_poll.h
 create mode 100644 lib/irq_poll.c

diff --git a/Documentation/kernel-per-CPU-kthreads.txt b/Documentation/kernel-per-CPU-kthreads.txt
index f4cbfe0..edec3a3 100644
--- a/Documentation/kernel-per-CPU-kthreads.txt
+++ b/Documentation/kernel-per-CPU-kthreads.txt
@@ -90,7 +90,7 @@ BLOCK_SOFTIRQ:  Do all of the following:
 	from being initiated from tasks that might run on the CPU to
 	be de-jittered.  (It is OK to force this CPU offline and then
 	bring it back online before you start your application.)
-BLOCK_IOPOLL_SOFTIRQ:  Do all of the following:
+IRQ_POLL_SOFTIRQ:  Do all of the following:
 1.	Force block-device interrupts onto some other CPU.
 2.	Initiate any block I/O and block-I/O polling on other CPUs.
 3.	Once your application has started, prevent CPU-hotplug operations
diff --git a/block/Makefile b/block/Makefile
index 00ecc97..e850474 100644
--- a/block/Makefile
+++ b/block/Makefile
@@ -5,7 +5,7 @@
 obj-$(CONFIG_BLOCK) := bio.o elevator.o blk-core.o blk-tag.o blk-sysfs.o \
 			blk-flush.o blk-settings.o blk-ioc.o blk-map.o \
 			blk-exec.o blk-merge.o blk-softirq.o blk-timeout.o \
-			blk-iopoll.o blk-lib.o blk-mq.o blk-mq-tag.o \
+			blk-lib.o blk-mq.o blk-mq-tag.o \
 			blk-mq-sysfs.o blk-mq-cpu.o blk-mq-cpumap.o ioctl.o \
 			genhd.o scsi_ioctl.o partition-generic.o ioprio.o \
 			partitions/
diff --git a/block/blk-iopoll.c b/block/blk-iopoll.c
deleted file mode 100644
index 0736729..0000000
--- a/block/blk-iopoll.c
+++ /dev/null
@@ -1,224 +0,0 @@
-/*
- * Functions related to interrupt-poll handling in the block layer. This
- * is similar to NAPI for network devices.
- */
-#include <linux/kernel.h>
-#include <linux/module.h>
-#include <linux/init.h>
-#include <linux/bio.h>
-#include <linux/blkdev.h>
-#include <linux/interrupt.h>
-#include <linux/cpu.h>
-#include <linux/blk-iopoll.h>
-#include <linux/delay.h>
-
-#include "blk.h"
-
-static unsigned int blk_iopoll_budget __read_mostly = 256;
-
-static DEFINE_PER_CPU(struct list_head, blk_cpu_iopoll);
-
-/**
- * blk_iopoll_sched - Schedule a run of the iopoll handler
- * @iop:      The parent iopoll structure
- *
- * Description:
- *     Add this blk_iopoll structure to the pending poll list and trigger the
- *     raise of the blk iopoll softirq. The driver must already have gotten a
- *     successful return from blk_iopoll_sched_prep() before calling this.
- **/
-void blk_iopoll_sched(struct blk_iopoll *iop)
-{
-	unsigned long flags;
-
-	local_irq_save(flags);
-	list_add_tail(&iop->list, this_cpu_ptr(&blk_cpu_iopoll));
-	__raise_softirq_irqoff(BLOCK_IOPOLL_SOFTIRQ);
-	local_irq_restore(flags);
-}
-EXPORT_SYMBOL(blk_iopoll_sched);
-
-/**
- * __blk_iopoll_complete - Mark this @iop as un-polled again
- * @iop:      The parent iopoll structure
- *
- * Description:
- *     See blk_iopoll_complete(). This function must be called with interrupts
- *     disabled.
- **/
-void __blk_iopoll_complete(struct blk_iopoll *iop)
-{
-	list_del(&iop->list);
-	smp_mb__before_atomic();
-	clear_bit_unlock(IOPOLL_F_SCHED, &iop->state);
-}
-EXPORT_SYMBOL(__blk_iopoll_complete);
-
-/**
- * blk_iopoll_complete - Mark this @iop as un-polled again
- * @iop:      The parent iopoll structure
- *
- * Description:
- *     If a driver consumes less than the assigned budget in its run of the
- *     iopoll handler, it'll end the polled mode by calling this function. The
- *     iopoll handler will not be invoked again before blk_iopoll_sched_prep()
- *     is called.
- **/
-void blk_iopoll_complete(struct blk_iopoll *iop)
-{
-	unsigned long flags;
-
-	local_irq_save(flags);
-	__blk_iopoll_complete(iop);
-	local_irq_restore(flags);
-}
-EXPORT_SYMBOL(blk_iopoll_complete);
-
-static void blk_iopoll_softirq(struct softirq_action *h)
-{
-	struct list_head *list = this_cpu_ptr(&blk_cpu_iopoll);
-	int rearm = 0, budget = blk_iopoll_budget;
-	unsigned long start_time = jiffies;
-
-	local_irq_disable();
-
-	while (!list_empty(list)) {
-		struct blk_iopoll *iop;
-		int work, weight;
-
-		/*
-		 * If softirq window is exhausted then punt.
-		 */
-		if (budget <= 0 || time_after(jiffies, start_time)) {
-			rearm = 1;
-			break;
-		}
-
-		local_irq_enable();
-
-		/* Even though interrupts have been re-enabled, this
-		 * access is safe because interrupts can only add new
-		 * entries to the tail of this list, and only ->poll()
-		 * calls can remove this head entry from the list.
-		 */
-		iop = list_entry(list->next, struct blk_iopoll, list);
-
-		weight = iop->weight;
-		work = 0;
-		if (test_bit(IOPOLL_F_SCHED, &iop->state))
-			work = iop->poll(iop, weight);
-
-		budget -= work;
-
-		local_irq_disable();
-
-		/*
-		 * Drivers must not modify the iopoll state, if they
-		 * consume their assigned weight (or more, some drivers can't
-		 * easily just stop processing, they have to complete an
-		 * entire mask of commands).In such cases this code
-		 * still "owns" the iopoll instance and therefore can
-		 * move the instance around on the list at-will.
-		 */
-		if (work >= weight) {
-			if (blk_iopoll_disable_pending(iop))
-				__blk_iopoll_complete(iop);
-			else
-				list_move_tail(&iop->list, list);
-		}
-	}
-
-	if (rearm)
-		__raise_softirq_irqoff(BLOCK_IOPOLL_SOFTIRQ);
-
-	local_irq_enable();
-}
-
-/**
- * blk_iopoll_disable - Disable iopoll on this @iop
- * @iop:      The parent iopoll structure
- *
- * Description:
- *     Disable io polling and wait for any pending callbacks to have completed.
- **/
-void blk_iopoll_disable(struct blk_iopoll *iop)
-{
-	set_bit(IOPOLL_F_DISABLE, &iop->state);
-	while (test_and_set_bit(IOPOLL_F_SCHED, &iop->state))
-		msleep(1);
-	clear_bit(IOPOLL_F_DISABLE, &iop->state);
-}
-EXPORT_SYMBOL(blk_iopoll_disable);
-
-/**
- * blk_iopoll_enable - Enable iopoll on this @iop
- * @iop:      The parent iopoll structure
- *
- * Description:
- *     Enable iopoll on this @iop. Note that the handler run will not be
- *     scheduled, it will only mark it as active.
- **/
-void blk_iopoll_enable(struct blk_iopoll *iop)
-{
-	BUG_ON(!test_bit(IOPOLL_F_SCHED, &iop->state));
-	smp_mb__before_atomic();
-	clear_bit_unlock(IOPOLL_F_SCHED, &iop->state);
-}
-EXPORT_SYMBOL(blk_iopoll_enable);
-
-/**
- * blk_iopoll_init - Initialize this @iop
- * @iop:      The parent iopoll structure
- * @weight:   The default weight (or command completion budget)
- * @poll_fn:  The handler to invoke
- *
- * Description:
- *     Initialize this blk_iopoll structure. Before being actively used, the
- *     driver must call blk_iopoll_enable().
- **/
-void blk_iopoll_init(struct blk_iopoll *iop, int weight, blk_iopoll_fn *poll_fn)
-{
-	memset(iop, 0, sizeof(*iop));
-	INIT_LIST_HEAD(&iop->list);
-	iop->weight = weight;
-	iop->poll = poll_fn;
-	set_bit(IOPOLL_F_SCHED, &iop->state);
-}
-EXPORT_SYMBOL(blk_iopoll_init);
-
-static int blk_iopoll_cpu_notify(struct notifier_block *self,
-				 unsigned long action, void *hcpu)
-{
-	/*
-	 * If a CPU goes away, splice its entries to the current CPU
-	 * and trigger a run of the softirq
-	 */
-	if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) {
-		int cpu = (unsigned long) hcpu;
-
-		local_irq_disable();
-		list_splice_init(&per_cpu(blk_cpu_iopoll, cpu),
-				 this_cpu_ptr(&blk_cpu_iopoll));
-		__raise_softirq_irqoff(BLOCK_IOPOLL_SOFTIRQ);
-		local_irq_enable();
-	}
-
-	return NOTIFY_OK;
-}
-
-static struct notifier_block blk_iopoll_cpu_notifier = {
-	.notifier_call	= blk_iopoll_cpu_notify,
-};
-
-static __init int blk_iopoll_setup(void)
-{
-	int i;
-
-	for_each_possible_cpu(i)
-		INIT_LIST_HEAD(&per_cpu(blk_cpu_iopoll, i));
-
-	open_softirq(BLOCK_IOPOLL_SOFTIRQ, blk_iopoll_softirq);
-	register_hotcpu_notifier(&blk_iopoll_cpu_notifier);
-	return 0;
-}
-subsys_initcall(blk_iopoll_setup);
diff --git a/drivers/scsi/Kconfig b/drivers/scsi/Kconfig
index 5f692ae..9209965 100644
--- a/drivers/scsi/Kconfig
+++ b/drivers/scsi/Kconfig
@@ -1102,6 +1102,7 @@ config SCSI_IPR
 	tristate "IBM Power Linux RAID adapter support"
 	depends on PCI && SCSI && ATA
 	select FW_LOADER
+	select IRQ_POLL
 	---help---
 	  This driver supports the IBM Power Linux family RAID adapters.
 	  This includes IBM pSeries 5712, 5703, 5709, and 570A, as well
diff --git a/drivers/scsi/be2iscsi/Kconfig b/drivers/scsi/be2iscsi/Kconfig
index 4e7cad2..bad5f32 100644
--- a/drivers/scsi/be2iscsi/Kconfig
+++ b/drivers/scsi/be2iscsi/Kconfig
@@ -3,6 +3,7 @@ config BE2ISCSI
 	depends on PCI && SCSI && NET
 	select SCSI_ISCSI_ATTRS
 	select ISCSI_BOOT_SYSFS
+	select IRQ_POLL
 
 	help
 	This driver implements the iSCSI functionality for Emulex
diff --git a/drivers/scsi/be2iscsi/be.h b/drivers/scsi/be2iscsi/be.h
index 77f992e..a41c643 100644
--- a/drivers/scsi/be2iscsi/be.h
+++ b/drivers/scsi/be2iscsi/be.h
@@ -20,7 +20,7 @@
 
 #include <linux/pci.h>
 #include <linux/if_vlan.h>
-#include <linux/blk-iopoll.h>
+#include <linux/irq_poll.h>
 #define FW_VER_LEN	32
 #define MCC_Q_LEN	128
 #define MCC_CQ_LEN	256
@@ -101,7 +101,7 @@ struct be_eq_obj {
 	struct beiscsi_hba *phba;
 	struct be_queue_info *cq;
 	struct work_struct work_cqs; /* Work Item */
-	struct blk_iopoll	iopoll;
+	struct irq_poll	iopoll;
 };
 
 struct be_mcc_obj {
diff --git a/drivers/scsi/be2iscsi/be_iscsi.c b/drivers/scsi/be2iscsi/be_iscsi.c
index b7087ba..022e87b 100644
--- a/drivers/scsi/be2iscsi/be_iscsi.c
+++ b/drivers/scsi/be2iscsi/be_iscsi.c
@@ -1292,9 +1292,9 @@ static void beiscsi_flush_cq(struct beiscsi_hba *phba)
 
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_disable(&pbe_eq->iopoll);
+		irq_poll_disable(&pbe_eq->iopoll);
 		beiscsi_process_cq(pbe_eq);
-		blk_iopoll_enable(&pbe_eq->iopoll);
+		irq_poll_enable(&pbe_eq->iopoll);
 	}
 }
 
diff --git a/drivers/scsi/be2iscsi/be_main.c b/drivers/scsi/be2iscsi/be_main.c
index fe0c514..1d879ef 100644
--- a/drivers/scsi/be2iscsi/be_main.c
+++ b/drivers/scsi/be2iscsi/be_main.c
@@ -910,8 +910,8 @@ static irqreturn_t be_isr_msix(int irq, void *dev_id)
 	num_eq_processed = 0;
 	while (eqe->dw[offsetof(struct amap_eq_entry, valid) / 32]
 				& EQE_VALID_MASK) {
-		if (!blk_iopoll_sched_prep(&pbe_eq->iopoll))
-			blk_iopoll_sched(&pbe_eq->iopoll);
+		if (!irq_poll_sched_prep(&pbe_eq->iopoll))
+			irq_poll_sched(&pbe_eq->iopoll);
 
 		AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
 		queue_tail_inc(eq);
@@ -972,8 +972,8 @@ static irqreturn_t be_isr(int irq, void *dev_id)
 			spin_unlock_irqrestore(&phba->isr_lock, flags);
 			num_mcceq_processed++;
 		} else {
-			if (!blk_iopoll_sched_prep(&pbe_eq->iopoll))
-				blk_iopoll_sched(&pbe_eq->iopoll);
+			if (!irq_poll_sched_prep(&pbe_eq->iopoll))
+				irq_poll_sched(&pbe_eq->iopoll);
 			num_ioeq_processed++;
 		}
 		AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
@@ -2295,7 +2295,7 @@ void beiscsi_process_all_cqs(struct work_struct *work)
 	hwi_ring_eq_db(phba, pbe_eq->q.id, 0, 0, 1, 1);
 }
 
-static int be_iopoll(struct blk_iopoll *iop, int budget)
+static int be_iopoll(struct irq_poll *iop, int budget)
 {
 	unsigned int ret;
 	struct beiscsi_hba *phba;
@@ -2306,7 +2306,7 @@ static int be_iopoll(struct blk_iopoll *iop, int budget)
 	pbe_eq->cq_count += ret;
 	if (ret < budget) {
 		phba = pbe_eq->phba;
-		blk_iopoll_complete(iop);
+		irq_poll_complete(iop);
 		beiscsi_log(phba, KERN_INFO,
 			    BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
 			    "BM_%d : rearm pbe_eq->q.id =%d\n",
@@ -5293,7 +5293,7 @@ static void beiscsi_quiesce(struct beiscsi_hba *phba,
 
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_disable(&pbe_eq->iopoll);
+		irq_poll_disable(&pbe_eq->iopoll);
 	}
 
 	if (unload_state == BEISCSI_CLEAN_UNLOAD) {
@@ -5579,9 +5579,9 @@ static void beiscsi_eeh_resume(struct pci_dev *pdev)
 
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_init(&pbe_eq->iopoll, be_iopoll_budget,
+		irq_poll_init(&pbe_eq->iopoll, be_iopoll_budget,
 				be_iopoll);
-		blk_iopoll_enable(&pbe_eq->iopoll);
+		irq_poll_enable(&pbe_eq->iopoll);
 	}
 
 	i = (phba->msix_enabled) ? i : 0;
@@ -5752,9 +5752,9 @@ static int beiscsi_dev_probe(struct pci_dev *pcidev,
 
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_init(&pbe_eq->iopoll, be_iopoll_budget,
+		irq_poll_init(&pbe_eq->iopoll, be_iopoll_budget,
 				be_iopoll);
-		blk_iopoll_enable(&pbe_eq->iopoll);
+		irq_poll_enable(&pbe_eq->iopoll);
 	}
 
 	i = (phba->msix_enabled) ? i : 0;
@@ -5795,7 +5795,7 @@ free_blkenbld:
 	destroy_workqueue(phba->wq);
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_disable(&pbe_eq->iopoll);
+		irq_poll_disable(&pbe_eq->iopoll);
 	}
 free_twq:
 	beiscsi_clean_port(phba);
diff --git a/drivers/scsi/ipr.c b/drivers/scsi/ipr.c
index 536cd5a..6b9c738 100644
--- a/drivers/scsi/ipr.c
+++ b/drivers/scsi/ipr.c
@@ -3638,7 +3638,7 @@ static struct device_attribute ipr_ioa_reset_attr = {
 	.store = ipr_store_reset_adapter
 };
 
-static int ipr_iopoll(struct blk_iopoll *iop, int budget);
+static int ipr_iopoll(struct irq_poll *iop, int budget);
  /**
  * ipr_show_iopoll_weight - Show ipr polling mode
  * @dev:	class device struct
@@ -3681,34 +3681,34 @@ static ssize_t ipr_store_iopoll_weight(struct device *dev,
 	int i;
 
 	if (!ioa_cfg->sis64) {
-		dev_info(&ioa_cfg->pdev->dev, "blk-iopoll not supported on this adapter\n");
+		dev_info(&ioa_cfg->pdev->dev, "irq_poll not supported on this adapter\n");
 		return -EINVAL;
 	}
 	if (kstrtoul(buf, 10, &user_iopoll_weight))
 		return -EINVAL;
 
 	if (user_iopoll_weight > 256) {
-		dev_info(&ioa_cfg->pdev->dev, "Invalid blk-iopoll weight. It must be less than 256\n");
+		dev_info(&ioa_cfg->pdev->dev, "Invalid irq_poll weight. It must be less than 256\n");
 		return -EINVAL;
 	}
 
 	if (user_iopoll_weight == ioa_cfg->iopoll_weight) {
-		dev_info(&ioa_cfg->pdev->dev, "Current blk-iopoll weight has the same weight\n");
+		dev_info(&ioa_cfg->pdev->dev, "Current irq_poll weight has the same weight\n");
 		return strlen(buf);
 	}
 
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		for (i = 1; i < ioa_cfg->hrrq_num; i++)
-			blk_iopoll_disable(&ioa_cfg->hrrq[i].iopoll);
+			irq_poll_disable(&ioa_cfg->hrrq[i].iopoll);
 	}
 
 	spin_lock_irqsave(shost->host_lock, lock_flags);
 	ioa_cfg->iopoll_weight = user_iopoll_weight;
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		for (i = 1; i < ioa_cfg->hrrq_num; i++) {
-			blk_iopoll_init(&ioa_cfg->hrrq[i].iopoll,
+			irq_poll_init(&ioa_cfg->hrrq[i].iopoll,
 					ioa_cfg->iopoll_weight, ipr_iopoll);
-			blk_iopoll_enable(&ioa_cfg->hrrq[i].iopoll);
+			irq_poll_enable(&ioa_cfg->hrrq[i].iopoll);
 		}
 	}
 	spin_unlock_irqrestore(shost->host_lock, lock_flags);
@@ -5569,7 +5569,7 @@ static int ipr_process_hrrq(struct ipr_hrr_queue *hrr_queue, int budget,
 	return num_hrrq;
 }
 
-static int ipr_iopoll(struct blk_iopoll *iop, int budget)
+static int ipr_iopoll(struct irq_poll *iop, int budget)
 {
 	struct ipr_ioa_cfg *ioa_cfg;
 	struct ipr_hrr_queue *hrrq;
@@ -5585,7 +5585,7 @@ static int ipr_iopoll(struct blk_iopoll *iop, int budget)
 	completed_ops = ipr_process_hrrq(hrrq, budget, &doneq);
 
 	if (completed_ops < budget)
-		blk_iopoll_complete(iop);
+		irq_poll_complete(iop);
 	spin_unlock_irqrestore(hrrq->lock, hrrq_flags);
 
 	list_for_each_entry_safe(ipr_cmd, temp, &doneq, queue) {
@@ -5693,8 +5693,8 @@ static irqreturn_t ipr_isr_mhrrq(int irq, void *devp)
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		if ((be32_to_cpu(*hrrq->hrrq_curr) & IPR_HRRQ_TOGGLE_BIT) ==
 		       hrrq->toggle_bit) {
-			if (!blk_iopoll_sched_prep(&hrrq->iopoll))
-				blk_iopoll_sched(&hrrq->iopoll);
+			if (!irq_poll_sched_prep(&hrrq->iopoll))
+				irq_poll_sched(&hrrq->iopoll);
 			spin_unlock_irqrestore(hrrq->lock, hrrq_flags);
 			return IRQ_HANDLED;
 		}
@@ -10405,9 +10405,9 @@ static int ipr_probe(struct pci_dev *pdev, const struct pci_device_id *dev_id)
 
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		for (i = 1; i < ioa_cfg->hrrq_num; i++) {
-			blk_iopoll_init(&ioa_cfg->hrrq[i].iopoll,
+			irq_poll_init(&ioa_cfg->hrrq[i].iopoll,
 					ioa_cfg->iopoll_weight, ipr_iopoll);
-			blk_iopoll_enable(&ioa_cfg->hrrq[i].iopoll);
+			irq_poll_enable(&ioa_cfg->hrrq[i].iopoll);
 		}
 	}
 
@@ -10436,7 +10436,7 @@ static void ipr_shutdown(struct pci_dev *pdev)
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		ioa_cfg->iopoll_weight = 0;
 		for (i = 1; i < ioa_cfg->hrrq_num; i++)
-			blk_iopoll_disable(&ioa_cfg->hrrq[i].iopoll);
+			irq_poll_disable(&ioa_cfg->hrrq[i].iopoll);
 	}
 
 	while (ioa_cfg->in_reset_reload) {
diff --git a/drivers/scsi/ipr.h b/drivers/scsi/ipr.h
index a34c7a5..56c5706 100644
--- a/drivers/scsi/ipr.h
+++ b/drivers/scsi/ipr.h
@@ -32,7 +32,7 @@
 #include <linux/libata.h>
 #include <linux/list.h>
 #include <linux/kref.h>
-#include <linux/blk-iopoll.h>
+#include <linux/irq_poll.h>
 #include <scsi/scsi.h>
 #include <scsi/scsi_cmnd.h>
 
@@ -517,7 +517,7 @@ struct ipr_hrr_queue {
 	u8 allow_cmds:1;
 	u8 removing_ioa:1;
 
-	struct blk_iopoll iopoll;
+	struct irq_poll iopoll;
 };
 
 /* Command packet structure */
diff --git a/include/linux/blk-iopoll.h b/include/linux/blk-iopoll.h
deleted file mode 100644
index 77ae77c..0000000
--- a/include/linux/blk-iopoll.h
+++ /dev/null
@@ -1,46 +0,0 @@
-#ifndef BLK_IOPOLL_H
-#define BLK_IOPOLL_H
-
-struct blk_iopoll;
-typedef int (blk_iopoll_fn)(struct blk_iopoll *, int);
-
-struct blk_iopoll {
-	struct list_head list;
-	unsigned long state;
-	unsigned long data;
-	int weight;
-	int max;
-	blk_iopoll_fn *poll;
-};
-
-enum {
-	IOPOLL_F_SCHED		= 0,
-	IOPOLL_F_DISABLE	= 1,
-};
-
-/*
- * Returns 0 if we successfully set the IOPOLL_F_SCHED bit, indicating
- * that we were the first to acquire this iop for scheduling. If this iop
- * is currently disabled, return "failure".
- */
-static inline int blk_iopoll_sched_prep(struct blk_iopoll *iop)
-{
-	if (!test_bit(IOPOLL_F_DISABLE, &iop->state))
-		return test_and_set_bit(IOPOLL_F_SCHED, &iop->state);
-
-	return 1;
-}
-
-static inline int blk_iopoll_disable_pending(struct blk_iopoll *iop)
-{
-	return test_bit(IOPOLL_F_DISABLE, &iop->state);
-}
-
-extern void blk_iopoll_sched(struct blk_iopoll *);
-extern void blk_iopoll_init(struct blk_iopoll *, int, blk_iopoll_fn *);
-extern void blk_iopoll_complete(struct blk_iopoll *);
-extern void __blk_iopoll_complete(struct blk_iopoll *);
-extern void blk_iopoll_enable(struct blk_iopoll *);
-extern void blk_iopoll_disable(struct blk_iopoll *);
-
-#endif
diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h
index ad16809..7ff98c2 100644
--- a/include/linux/interrupt.h
+++ b/include/linux/interrupt.h
@@ -412,7 +412,7 @@ enum
 	NET_TX_SOFTIRQ,
 	NET_RX_SOFTIRQ,
 	BLOCK_SOFTIRQ,
-	BLOCK_IOPOLL_SOFTIRQ,
+	IRQ_POLL_SOFTIRQ,
 	TASKLET_SOFTIRQ,
 	SCHED_SOFTIRQ,
 	HRTIMER_SOFTIRQ, /* Unused, but kept as tools rely on the
diff --git a/include/linux/irq_poll.h b/include/linux/irq_poll.h
new file mode 100644
index 0000000..0cf7c26
--- /dev/null
+++ b/include/linux/irq_poll.h
@@ -0,0 +1,46 @@
+#ifndef IRQ_POLL_H
+#define IRQ_POLL_H
+
+struct irq_poll;
+typedef int (irq_poll_fn)(struct irq_poll *, int);
+
+struct irq_poll {
+	struct list_head list;
+	unsigned long state;
+	unsigned long data;
+	int weight;
+	int max;
+	irq_poll_fn *poll;
+};
+
+enum {
+	IRQ_POLL_F_SCHED		= 0,
+	IRQ_POLL_F_DISABLE	= 1,
+};
+
+/*
+ * Returns 0 if we successfully set the IRQ_POLL_F_SCHED bit, indicating
+ * that we were the first to acquire this iop for scheduling. If this iop
+ * is currently disabled, return "failure".
+ */
+static inline int irq_poll_sched_prep(struct irq_poll *iop)
+{
+	if (!test_bit(IRQ_POLL_F_DISABLE, &iop->state))
+		return test_and_set_bit(IRQ_POLL_F_SCHED, &iop->state);
+
+	return 1;
+}
+
+static inline int irq_poll_disable_pending(struct irq_poll *iop)
+{
+	return test_bit(IRQ_POLL_F_DISABLE, &iop->state);
+}
+
+extern void irq_poll_sched(struct irq_poll *);
+extern void irq_poll_init(struct irq_poll *, int, irq_poll_fn *);
+extern void irq_poll_complete(struct irq_poll *);
+extern void __irq_poll_complete(struct irq_poll *);
+extern void irq_poll_enable(struct irq_poll *);
+extern void irq_poll_disable(struct irq_poll *);
+
+#endif
diff --git a/include/trace/events/irq.h b/include/trace/events/irq.h
index ff8f6c0..f95f25e 100644
--- a/include/trace/events/irq.h
+++ b/include/trace/events/irq.h
@@ -15,7 +15,7 @@ struct softirq_action;
 			 softirq_name(NET_TX)		\
 			 softirq_name(NET_RX)		\
 			 softirq_name(BLOCK)		\
-			 softirq_name(BLOCK_IOPOLL)	\
+			 softirq_name(IRQ_POLL)		\
 			 softirq_name(TASKLET)		\
 			 softirq_name(SCHED)		\
 			 softirq_name(HRTIMER)		\
diff --git a/lib/Kconfig b/lib/Kconfig
index f0df318..e00e196 100644
--- a/lib/Kconfig
+++ b/lib/Kconfig
@@ -475,6 +475,11 @@ config DDR
 	  information. This data is useful for drivers handling
 	  DDR SDRAM controllers.
 
+config IRQ_POLL
+	bool "IRQ polling library"
+	help
+	  Helper library to poll interrupt mitigation using polling.
+
 config MPILIB
 	tristate
 	select CLZ_TAB
diff --git a/lib/Makefile b/lib/Makefile
index 7f1de26..1478ae2 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -164,6 +164,7 @@ obj-$(CONFIG_GENERIC_NET_UTILS) += net_utils.o
 
 obj-$(CONFIG_SG_SPLIT) += sg_split.o
 obj-$(CONFIG_STMP_DEVICE) += stmp_device.o
+obj-$(CONFIG_IRQ_POLL) += irq_poll.o
 
 libfdt_files = fdt.o fdt_ro.o fdt_wip.o fdt_rw.o fdt_sw.o fdt_strerror.o \
 	       fdt_empty_tree.o
diff --git a/lib/irq_poll.c b/lib/irq_poll.c
new file mode 100644
index 0000000..e6fd1dc
--- /dev/null
+++ b/lib/irq_poll.c
@@ -0,0 +1,221 @@
+/*
+ * Functions related to interrupt-poll handling in the block layer. This
+ * is similar to NAPI for network devices.
+ */
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/bio.h>
+#include <linux/interrupt.h>
+#include <linux/cpu.h>
+#include <linux/irq_poll.h>
+#include <linux/delay.h>
+
+static unsigned int irq_poll_budget __read_mostly = 256;
+
+static DEFINE_PER_CPU(struct list_head, blk_cpu_iopoll);
+
+/**
+ * irq_poll_sched - Schedule a run of the iopoll handler
+ * @iop:      The parent iopoll structure
+ *
+ * Description:
+ *     Add this irq_poll structure to the pending poll list and trigger the
+ *     raise of the blk iopoll softirq. The driver must already have gotten a
+ *     successful return from irq_poll_sched_prep() before calling this.
+ **/
+void irq_poll_sched(struct irq_poll *iop)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+	list_add_tail(&iop->list, this_cpu_ptr(&blk_cpu_iopoll));
+	__raise_softirq_irqoff(IRQ_POLL_SOFTIRQ);
+	local_irq_restore(flags);
+}
+EXPORT_SYMBOL(irq_poll_sched);
+
+/**
+ * __irq_poll_complete - Mark this @iop as un-polled again
+ * @iop:      The parent iopoll structure
+ *
+ * Description:
+ *     See irq_poll_complete(). This function must be called with interrupts
+ *     disabled.
+ **/
+void __irq_poll_complete(struct irq_poll *iop)
+{
+	list_del(&iop->list);
+	smp_mb__before_atomic();
+	clear_bit_unlock(IRQ_POLL_F_SCHED, &iop->state);
+}
+EXPORT_SYMBOL(__irq_poll_complete);
+
+/**
+ * irq_poll_complete - Mark this @iop as un-polled again
+ * @iop:      The parent iopoll structure
+ *
+ * Description:
+ *     If a driver consumes less than the assigned budget in its run of the
+ *     iopoll handler, it'll end the polled mode by calling this function. The
+ *     iopoll handler will not be invoked again before irq_poll_sched_prep()
+ *     is called.
+ **/
+void irq_poll_complete(struct irq_poll *iop)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+	__irq_poll_complete(iop);
+	local_irq_restore(flags);
+}
+EXPORT_SYMBOL(irq_poll_complete);
+
+static void irq_poll_softirq(struct softirq_action *h)
+{
+	struct list_head *list = this_cpu_ptr(&blk_cpu_iopoll);
+	int rearm = 0, budget = irq_poll_budget;
+	unsigned long start_time = jiffies;
+
+	local_irq_disable();
+
+	while (!list_empty(list)) {
+		struct irq_poll *iop;
+		int work, weight;
+
+		/*
+		 * If softirq window is exhausted then punt.
+		 */
+		if (budget <= 0 || time_after(jiffies, start_time)) {
+			rearm = 1;
+			break;
+		}
+
+		local_irq_enable();
+
+		/* Even though interrupts have been re-enabled, this
+		 * access is safe because interrupts can only add new
+		 * entries to the tail of this list, and only ->poll()
+		 * calls can remove this head entry from the list.
+		 */
+		iop = list_entry(list->next, struct irq_poll, list);
+
+		weight = iop->weight;
+		work = 0;
+		if (test_bit(IRQ_POLL_F_SCHED, &iop->state))
+			work = iop->poll(iop, weight);
+
+		budget -= work;
+
+		local_irq_disable();
+
+		/*
+		 * Drivers must not modify the iopoll state, if they
+		 * consume their assigned weight (or more, some drivers can't
+		 * easily just stop processing, they have to complete an
+		 * entire mask of commands).In such cases this code
+		 * still "owns" the iopoll instance and therefore can
+		 * move the instance around on the list at-will.
+		 */
+		if (work >= weight) {
+			if (irq_poll_disable_pending(iop))
+				__irq_poll_complete(iop);
+			else
+				list_move_tail(&iop->list, list);
+		}
+	}
+
+	if (rearm)
+		__raise_softirq_irqoff(IRQ_POLL_SOFTIRQ);
+
+	local_irq_enable();
+}
+
+/**
+ * irq_poll_disable - Disable iopoll on this @iop
+ * @iop:      The parent iopoll structure
+ *
+ * Description:
+ *     Disable io polling and wait for any pending callbacks to have completed.
+ **/
+void irq_poll_disable(struct irq_poll *iop)
+{
+	set_bit(IRQ_POLL_F_DISABLE, &iop->state);
+	while (test_and_set_bit(IRQ_POLL_F_SCHED, &iop->state))
+		msleep(1);
+	clear_bit(IRQ_POLL_F_DISABLE, &iop->state);
+}
+EXPORT_SYMBOL(irq_poll_disable);
+
+/**
+ * irq_poll_enable - Enable iopoll on this @iop
+ * @iop:      The parent iopoll structure
+ *
+ * Description:
+ *     Enable iopoll on this @iop. Note that the handler run will not be
+ *     scheduled, it will only mark it as active.
+ **/
+void irq_poll_enable(struct irq_poll *iop)
+{
+	BUG_ON(!test_bit(IRQ_POLL_F_SCHED, &iop->state));
+	smp_mb__before_atomic();
+	clear_bit_unlock(IRQ_POLL_F_SCHED, &iop->state);
+}
+EXPORT_SYMBOL(irq_poll_enable);
+
+/**
+ * irq_poll_init - Initialize this @iop
+ * @iop:      The parent iopoll structure
+ * @weight:   The default weight (or command completion budget)
+ * @poll_fn:  The handler to invoke
+ *
+ * Description:
+ *     Initialize this irq_poll structure. Before being actively used, the
+ *     driver must call irq_poll_enable().
+ **/
+void irq_poll_init(struct irq_poll *iop, int weight, irq_poll_fn *poll_fn)
+{
+	memset(iop, 0, sizeof(*iop));
+	INIT_LIST_HEAD(&iop->list);
+	iop->weight = weight;
+	iop->poll = poll_fn;
+	set_bit(IRQ_POLL_F_SCHED, &iop->state);
+}
+EXPORT_SYMBOL(irq_poll_init);
+
+static int irq_poll_cpu_notify(struct notifier_block *self,
+				 unsigned long action, void *hcpu)
+{
+	/*
+	 * If a CPU goes away, splice its entries to the current CPU
+	 * and trigger a run of the softirq
+	 */
+	if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) {
+		int cpu = (unsigned long) hcpu;
+
+		local_irq_disable();
+		list_splice_init(&per_cpu(blk_cpu_iopoll, cpu),
+				 this_cpu_ptr(&blk_cpu_iopoll));
+		__raise_softirq_irqoff(IRQ_POLL_SOFTIRQ);
+		local_irq_enable();
+	}
+
+	return NOTIFY_OK;
+}
+
+static struct notifier_block irq_poll_cpu_notifier = {
+	.notifier_call	= irq_poll_cpu_notify,
+};
+
+static __init int irq_poll_setup(void)
+{
+	int i;
+
+	for_each_possible_cpu(i)
+		INIT_LIST_HEAD(&per_cpu(blk_cpu_iopoll, i));
+
+	open_softirq(IRQ_POLL_SOFTIRQ, irq_poll_softirq);
+	register_hotcpu_notifier(&irq_poll_cpu_notifier);
+	return 0;
+}
+subsys_initcall(irq_poll_setup);
diff --git a/tools/lib/traceevent/event-parse.c b/tools/lib/traceevent/event-parse.c
index 2a912df..af5a316 100644
--- a/tools/lib/traceevent/event-parse.c
+++ b/tools/lib/traceevent/event-parse.c
@@ -3746,7 +3746,7 @@ static const struct flag flags[] = {
 	{ "NET_TX_SOFTIRQ", 2 },
 	{ "NET_RX_SOFTIRQ", 3 },
 	{ "BLOCK_SOFTIRQ", 4 },
-	{ "BLOCK_IOPOLL_SOFTIRQ", 5 },
+	{ "IRQ_POLL_SOFTIRQ", 5 },
 	{ "TASKLET_SOFTIRQ", 6 },
 	{ "SCHED_SOFTIRQ", 7 },
 	{ "HRTIMER_SOFTIRQ", 8 },
diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c
index 8ff7d62..33b52ea 100644
--- a/tools/perf/util/trace-event-parse.c
+++ b/tools/perf/util/trace-event-parse.c
@@ -209,7 +209,7 @@ static const struct flag flags[] = {
 	{ "NET_TX_SOFTIRQ", 2 },
 	{ "NET_RX_SOFTIRQ", 3 },
 	{ "BLOCK_SOFTIRQ", 4 },
-	{ "BLOCK_IOPOLL_SOFTIRQ", 5 },
+	{ "IRQ_POLL_SOFTIRQ", 5 },
 	{ "TASKLET_SOFTIRQ", 6 },
 	{ "SCHED_SOFTIRQ", 7 },
 	{ "HRTIMER_SOFTIRQ", 8 },
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web