Hello chenxiang,

On Tue, Feb 06, 2018 at 10:18:19AM +0800, chenxiang (M) wrote:

在 2018/2/5 23:20, Ming Lei 写道:

quoted

This patch uses .force_blk_mq to drive HPSA via SCSI_MQ, meantime maps
each reply queue to blk_mq's hw queue, then .queuecommand can always
choose the hw queue as the reply queue. And if no any online CPU is
mapped to one hw queue, request can't be submitted to this hw queue
at all, finally the irq affinity issue is solved.

Cc: Hannes Reinecke <hare@suse.de>
Cc: Arun Easi <redacted>
Cc: Omar Sandoval <redacted>,
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>,
Cc: James Bottomley <james.bottomley@hansenpartnership.com>,
Cc: Christoph Hellwig <hch@lst.de>,
Cc: Don Brace <redacted>
Cc: Kashyap Desai <kashyap.desai@broadcom.com>
Cc: Peter Rivera <redacted>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Mike Snitzer <redacted>
Tested-by: Laurence Oberman <redacted>
Signed-off-by: Ming Lei <redacted>
---
  drivers/scsi/hpsa.c | 51 ++++++++++++++++++++++++++++++++++-----------------
  1 file changed, 34 insertions(+), 17 deletions(-)

diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c
index 443eabf63a9f..e517a4c74a28 100644
--- a/drivers/scsi/hpsa.c
+++ b/drivers/scsi/hpsa.c

@@ -51,6 +51,7 @@
  #include <linux/jiffies.h>
  #include <linux/percpu-defs.h>
  #include <linux/percpu.h>
+#include <linux/blk-mq-pci.h>
  #include <asm/unaligned.h>
  #include <asm/div64.h>
  #include "hpsa_cmd.h"

@@ -956,6 +957,13 @@ static struct device_attribute *hpsa_shost_attrs[] = {
  #define HPSA_NRESERVED_CMDS	(HPSA_CMDS_RESERVED_FOR_DRIVER +\
  				 HPSA_MAX_CONCURRENT_PASSTHRUS)
+static int hpsa_map_queues(struct Scsi_Host *shost)
+{
+        struct ctlr_info *h = shost_to_hba(shost);
+
+        return blk_mq_pci_map_queues(&shost->tag_set, h->pdev);
+}
+

Hi Lei Ming,
It is okay to use blk_mq_pci_map_queue to solve automatic irq affinity issue
when the first interrupt vector for queues is 0.
But if the first interrupt vector for queues is not 0,  we seems couldn't
use blk_mq_pci_map_queue directly,
such as blk_mq_virtio_map_queues, it realizes a interface itself. Is it
possible to provide a general interface for those
situations?

I guess it isn't necessary to do that, as you see .map_queues has been
introduced to 'scsi_host_template' for dealing driver specific irq
vector difference, such as, virtio-pci, 'irq_affinity' is needed for
excluding 'pre_vectors' which should serve as virtio config vector.

But that should belong to another topic about implementing generic
.map_queues interface, and seems not related with this patch, since
the usage of blk_mq_pci_map_queues() in this patch is correct.

Thanks,
Ming