IB/srp: Add multichannel support

Improve performance by using multiple RDMA/RC channels per SCSI
host for communication with an SRP target. About the
implementation:
- Introduce a loop over all channels in the code that uses
  target->ch.
- Set the SRP_MULTICHAN_MULTI flag during login for the creation
  of the second and subsequent channels.
- RDMA completion vectors are chosen such that RDMA completion
  interrupts are handled by the CPU socket that submitted the I/O
  request. As one can see in this patch it has been assumed if a
  system contains n CPU sockets and m RDMA completion vectors
  have been assigned to an RDMA HCA that IRQ affinity has been
  configured such that completion vectors [i*m/n..(i+1)*m/n) are
  bound to CPU socket i with 0 <= i < n.
- Modify srp_free_ch_ib() and srp_free_req_data() such that it
  becomes safe to invoke these functions after the corresponding
  allocation function failed.
- Add a ch_count sysfs attribute per target port.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

Documentation/ABI/stable/sysfs-driver-ib_srp

@@ -55,12 +55,12 @@ Description:	Interface for making ib_srp connect to a new target.
 		  only safe with partial memory descriptor list support enabled
 		  (allow_ext_sg=1).
 		* comp_vector, a number in the range 0..n-1 specifying the
-		  MSI-X completion vector. Some HCA's allocate multiple (n)
-		  MSI-X vectors per HCA port. If the IRQ affinity masks of
-		  these interrupts have been configured such that each MSI-X
-		  interrupt is handled by a different CPU then the comp_vector
-		  parameter can be used to spread the SRP completion workload
-		  over multiple CPU's.
+		  MSI-X completion vector of the first RDMA channel. Some
+		  HCA's allocate multiple (n) MSI-X vectors per HCA port. If
+		  the IRQ affinity masks of these interrupts have been
+		  configured such that each MSI-X interrupt is handled by a
+		  different CPU then the comp_vector parameter can be used to
+		  spread the SRP completion workload over multiple CPU's.
 		* tl_retry_count, a number in the range 2..7 specifying the
 		  IB RC retry count.
 		* queue_size, the maximum number of commands that the
@@ -88,6 +88,13 @@ Description:	Whether ib_srp is allowed to include a partial memory
 		descriptor list in an SRP_CMD when communicating with an SRP
 		target.
 
+What:		/sys/class/scsi_host/host<n>/ch_count
+Date:		April 1, 2015
+KernelVersion:	3.19
+Contact:	linux-rdma@vger.kernel.org
+Description:	Number of RDMA channels used for communication with the SRP
+		target.
+
 What:		/sys/class/scsi_host/host<n>/cmd_sg_entries
 Date:		May 19, 2011
 KernelVersion:	2.6.39
@@ -95,6 +102,12 @@ Contact:	linux-rdma@vger.kernel.org
 Description:	Maximum number of data buffer descriptors that may be sent to
 		the target in a single SRP_CMD request.
 
+What:		/sys/class/scsi_host/host<n>/comp_vector
+Date:		September 2, 2013
+KernelVersion:	3.11
+Contact:	linux-rdma@vger.kernel.org
+Description:	Completion vector used for the first RDMA channel.
+
 What:		/sys/class/scsi_host/host<n>/dgid
 Date:		June 17, 2006
 KernelVersion:	2.6.17

drivers/infiniband/ulp/srp/ib_srp.h

@@ -179,8 +179,9 @@ struct srp_target_port {
 	/* read and written in the hot path */
 	spinlock_t		lock;
 
-	struct srp_rdma_ch	ch;
 	/* read only in the hot path */
+	struct srp_rdma_ch	*ch;
+	u32			ch_count;
 	u32			lkey;
 	u32			rkey;
 	enum srp_target_state	state;