We always assign a dummy task context to a port in order to address a silicon issue. We have 4 ports per controller. So when idle, there are always exactly 4 TCs "active". The adaptive interrupt coalescing code uses number of active TCs to figure out the coalescing values. However, we never hit "0" TCs because of the 4 dummy TCs. Putting in fix so that we calculate this correctly.
Reported-by: Dan Melnic <[email protected]> Signed-off-by: Dave Jiang <[email protected]> --- drivers/scsi/isci/host.c | 10 ++++++++-- 1 files changed, 8 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/isci/host.c b/drivers/scsi/isci/host.c index b425ed5..bbe9fa5 100644 --- a/drivers/scsi/isci/host.c +++ b/drivers/scsi/isci/host.c @@ -1120,10 +1120,16 @@ void isci_host_completion_routine(unsigned long data) sci_controller_completion_handler(ihost); spin_unlock_irq(&ihost->scic_lock); - /* the coalesence timeout doubles at each encoding step, so + /* + * we subtract SCI_MAX_PORTS to account for the number of dummy TCs + * issued for hardware issue workaround + */ + active = isci_tci_active(ihost) - SCI_MAX_PORTS; + + /* + * the coalesence timeout doubles at each encoding step, so * update it based on the ilog2 value of the outstanding requests */ - active = isci_tci_active(ihost); writel(SMU_ICC_GEN_VAL(NUMBER, active) | SMU_ICC_GEN_VAL(TIMER, ISCI_COALESCE_BASE + ilog2(active)), &ihost->smu_registers->interrupt_coalesce_control); -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
