[SCSI] qla2xxx: Reduce request queue-size overhead with recent ISPs.
The original code to 'resize request-queues' based on iocb-counts
and employed during early ISP23xx testing was too
overly-pessimistic with regards to latencies in the firmware
pulling requests. Recent ISPs can easily keep up processing a
stream of commands from an abbreviated (effectively, half the
original size) queue.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
diff --git a/drivers/scsi/qla2xxx/qla_init.c b/drivers/scsi/qla2xxx/qla_init.c
index 997c83b..f90fd37 100644
--- a/drivers/scsi/qla2xxx/qla_init.c
+++ b/drivers/scsi/qla2xxx/qla_init.c
@@ -20,7 +20,6 @@
* QLogic ISP2x00 Hardware Support Function Prototypes.
*/
static int qla2x00_isp_firmware(scsi_qla_host_t *);
-static void qla2x00_resize_request_q(scsi_qla_host_t *);
static int qla2x00_setup_chip(scsi_qla_host_t *);
static int qla2x00_init_rings(scsi_qla_host_t *);
static int qla2x00_fw_ready(scsi_qla_host_t *);
@@ -893,62 +892,6 @@
}
/**
- * qla2x00_resize_request_q() - Resize request queue given available ISP memory.
- * @ha: HA context
- *
- * Returns 0 on success.
- */
-static void
-qla2x00_resize_request_q(scsi_qla_host_t *vha)
-{
- int rval;
- uint16_t fw_iocb_cnt = 0;
- uint16_t request_q_length = REQUEST_ENTRY_CNT_2XXX_EXT_MEM;
- dma_addr_t request_dma;
- request_t *request_ring;
- struct qla_hw_data *ha = vha->hw;
- struct req_que *req = ha->req_q_map[0];
-
- /* Valid only on recent ISPs. */
- if (IS_QLA2100(ha) || IS_QLA2200(ha))
- return;
-
- /* Retrieve IOCB counts available to the firmware. */
- rval = qla2x00_get_resource_cnts(vha, NULL, NULL, NULL, &fw_iocb_cnt,
- &ha->max_npiv_vports);
- if (rval)
- return;
- /* No point in continuing if current settings are sufficient. */
- if (fw_iocb_cnt < 1024)
- return;
- if (req->length >= request_q_length)
- return;
-
- /* Attempt to claim larger area for request queue. */
- request_ring = dma_alloc_coherent(&ha->pdev->dev,
- (request_q_length + 1) * sizeof(request_t), &request_dma,
- GFP_KERNEL);
- if (request_ring == NULL)
- return;
-
- /* Resize successful, report extensions. */
- qla_printk(KERN_INFO, ha, "Extended memory detected (%d KB)...\n",
- (ha->fw_memory_size + 1) / 1024);
- qla_printk(KERN_INFO, ha, "Resizing request queue depth "
- "(%d -> %d)...\n", req->length, request_q_length);
-
- /* Clear old allocations. */
- dma_free_coherent(&ha->pdev->dev,
- (req->length + 1) * sizeof(request_t), req->ring,
- req->dma);
-
- /* Begin using larger queue. */
- req->length = request_q_length;
- req->ring = request_ring;
- req->dma = request_dma;
-}
-
-/**
* qla2x00_setup_chip() - Load and start RISC firmware.
* @ha: HA context
*
@@ -1005,11 +948,11 @@
ha->max_npiv_vports =
MIN_MULTI_ID_FABRIC - 1;
}
- if (!fw_major_version) {
- qla2x00_resize_request_q(vha);
- if (ql2xallocfwdump)
- qla2x00_alloc_fw_dump(vha);
- }
+ qla2x00_get_resource_cnts(vha, NULL, NULL,
+ NULL, NULL, &ha->max_npiv_vports);
+
+ if (!fw_major_version && ql2xallocfwdump)
+ qla2x00_alloc_fw_dump(vha);
}
} else {
DEBUG2(printk(KERN_INFO