scsi: megaraid_sas: Introduce various Aero performance modes

For Aero adapters, driver provides three different performance modes
controlled through module parameter named 'perf_mode'. Below are those
performance modes:

 0: Balanced - Additional high IOPS reply queues will be enabled along with
    low latency queues. Interrupt coalescing will be enabled only for these
    high IOPS reply queues.

 1: IOPS - No additional high IOPS queues are enabled. Interrupt coalescing
    will be enabled on all reply queues.

 2: Latency - No additional high IOPS queues are enabled. Interrupt
    coalescing will be disabled on all reply queues. This is a legacy
    behavior similar to Ventura & Invader Series.

Default performance mode settings:

 - Performance mode set to 'Balanced', if Aero controller is working in
   16GT/s PCIe speed.

 - Performance mode will be set to 'Latency' mode for all other cases.

Through module parameter 'perf_mode', user can override default performance
mode to desired one.

Captured some performance numbers with these performance modes.  4k Random
Read IO performance numbers on 24 SAS SSD drives for above three
performance modes. Performance data is from Intel Skylake and HGST SS300
(drive model SDLL1DLR400GCCA1).

IOPS:
 -----------------------------------------------------------------------
  |perf_mode    | qd = 1 | qd = 64 |   note                             |
  |-------------|--------|---------|-------------------------------------
  |balanced     |  259K  |  3061k  | Provides max performance numbers   |
  |             |        |         | both on lower QD workload &        |
  |             |        |         | also on higher QD workload         |
  |-------------|--------|---------|-------------------------------------
  |iops         |  220K  |  3100k  | Provides max performance numbers   |
  |             |        |         | only on higher QD workload.        |
  |-------------|--------|---------|-------------------------------------
  |latency      |  246k  |  2226k  | Provides good performance numbers  |
  |             |        |         | only on lower QD worklaod.         |
  -----------------------------------------------------------------------

Average Latency:
  -----------------------------------------------------
  |perf_mode    |  qd = 1      |    qd = 64           |
  |-------------|--------------|----------------------|
  |balanced     |  92.05 usec  |    501.12 usec       |
  |-------------|--------------|----------------------|
  |iops         |  108.40 usec |    498.10 usec       |
  |-------------|--------------|----------------------|
  |latency      |  97.10 usec  |    689.26 usec       |
  -----------------------------------------------------

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Chandrakanth Patil <chandrakanth.patil@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This commit is contained in:
Chandrakanth Patil 2019-06-25 16:34:35 +05:30 committed by Martin K. Petersen
parent f39e5e52c5
commit 299ee42615
3 changed files with 81 additions and 15 deletions

View File

@ -2256,6 +2256,18 @@ enum MR_PD_TYPE {
#define MR_DEVICE_HIGH_IOPS_DEPTH 8
#define MR_HIGH_IOPS_BATCH_COUNT 16
enum MR_PERF_MODE {
MR_BALANCED_PERF_MODE = 0,
MR_IOPS_PERF_MODE = 1,
MR_LATENCY_PERF_MODE = 2,
};
#define MEGASAS_PERF_MODE_2STR(mode) \
((mode) == MR_BALANCED_PERF_MODE ? "Balanced" : \
(mode) == MR_IOPS_PERF_MODE ? "IOPS" : \
(mode) == MR_LATENCY_PERF_MODE ? "Latency" : \
"Unknown")
struct megasas_instance {
unsigned int *reply_map;
@ -2441,7 +2453,7 @@ struct megasas_instance {
bool support_seqnum_jbod_fp;
bool support_pci_lane_margining;
u8 low_latency_index_start;
bool balanced_mode;
int perf_mode;
};
struct MR_LD_VF_MAP {

View File

@ -105,6 +105,18 @@ unsigned int scmd_timeout = MEGASAS_DEFAULT_CMD_TIMEOUT;
module_param(scmd_timeout, int, 0444);
MODULE_PARM_DESC(scmd_timeout, "scsi command timeout (10-90s), default 90s. See megasas_reset_timer.");
int perf_mode = -1;
module_param(perf_mode, int, 0444);
MODULE_PARM_DESC(perf_mode, "Performance mode (only for Aero adapters), options:\n\t\t"
"0 - balanced: High iops and low latency queues are allocated &\n\t\t"
"interrupt coalescing is enabled only on high iops queues\n\t\t"
"1 - iops: High iops queues are not allocated &\n\t\t"
"interrupt coalescing is enabled on all queues\n\t\t"
"2 - latency: High iops queues are not allocated &\n\t\t"
"interrupt coalescing is disabled on all queues\n\t\t"
"default mode is 'balanced'"
);
MODULE_LICENSE("GPL");
MODULE_VERSION(MEGASAS_VERSION);
MODULE_AUTHOR("megaraidlinux.pdl@broadcom.com");
@ -5472,7 +5484,7 @@ megasas_setup_irqs_ioapic(struct megasas_instance *instance)
__func__, __LINE__);
return -1;
}
instance->balanced_mode = false;
instance->perf_mode = MR_LATENCY_PERF_MODE;
instance->low_latency_index_start = 0;
return 0;
}
@ -5683,7 +5695,7 @@ megasas_set_high_iops_queue_affinity_hint(struct megasas_instance *instance)
int i;
int local_numa_node;
if (instance->balanced_mode) {
if (instance->perf_mode == MR_BALANCED_PERF_MODE) {
local_numa_node = dev_to_node(&instance->pdev->dev);
for (i = 0; i < instance->low_latency_index_start; i++)
@ -5726,11 +5738,12 @@ megasas_alloc_irq_vectors(struct megasas_instance *instance)
i = __megasas_alloc_irq_vectors(instance);
if (instance->balanced_mode && (i != instance->msix_vectors)) {
if ((instance->perf_mode == MR_BALANCED_PERF_MODE) &&
(i != instance->msix_vectors)) {
if (instance->msix_vectors)
pci_free_irq_vectors(instance->pdev);
/* Disable Balanced IOPS mode and try realloc vectors */
instance->balanced_mode = false;
instance->perf_mode = MR_LATENCY_PERF_MODE;
instance->low_latency_index_start = 1;
num_msix_req = num_online_cpus() + instance->low_latency_index_start;
@ -5774,6 +5787,7 @@ static int megasas_init_fw(struct megasas_instance *instance)
struct fusion_context *fusion;
bool intr_coalescing;
unsigned int num_msix_req;
u16 lnksta, speed;
fusion = instance->ctrl_context;
@ -5983,11 +5997,43 @@ static int megasas_init_fw(struct megasas_instance *instance)
if (intr_coalescing &&
(num_online_cpus() >= MR_HIGH_IOPS_QUEUE_COUNT) &&
(instance->msix_vectors == MEGASAS_MAX_MSIX_QUEUES))
instance->balanced_mode = true;
instance->perf_mode = MR_BALANCED_PERF_MODE;
else
instance->balanced_mode = false;
instance->perf_mode = MR_LATENCY_PERF_MODE;
if (instance->balanced_mode)
if (instance->adapter_type == AERO_SERIES) {
pcie_capability_read_word(instance->pdev, PCI_EXP_LNKSTA, &lnksta);
speed = lnksta & PCI_EXP_LNKSTA_CLS;
/*
* For Aero, if PCIe link speed is <16 GT/s, then driver should operate
* in latency perf mode and enable R1 PCI bandwidth algorithm
*/
if (speed < 0x4) {
instance->perf_mode = MR_LATENCY_PERF_MODE;
fusion->pcie_bw_limitation = true;
}
/*
* Performance mode settings provided through module parameter-perf_mode will
* take affect only for:
* 1. Aero family of adapters.
* 2. When user sets module parameter- perf_mode in range of 0-2.
*/
if ((perf_mode >= MR_BALANCED_PERF_MODE) &&
(perf_mode <= MR_LATENCY_PERF_MODE))
instance->perf_mode = perf_mode;
/*
* If intr coalescing is not supported by controller FW, then IOPS
* and Balanced modes are not feasible.
*/
if (!intr_coalescing)
instance->perf_mode = MR_LATENCY_PERF_MODE;
}
if (instance->perf_mode == MR_BALANCED_PERF_MODE)
instance->low_latency_index_start =
MR_HIGH_IOPS_QUEUE_COUNT;
else

View File

@ -1097,10 +1097,10 @@ megasas_ioc_init_fusion(struct megasas_instance *instance)
if ((instance->low_latency_index_start ==
MR_HIGH_IOPS_QUEUE_COUNT) && cur_intr_coalescing)
instance->balanced_mode = true;
instance->perf_mode = MR_BALANCED_PERF_MODE;
dev_info(&instance->pdev->dev, "Balanced mode :%s\n",
instance->balanced_mode ? "Yes" : "No");
dev_info(&instance->pdev->dev, "Performance mode :%s\n",
MEGASAS_PERF_MODE_2STR(instance->perf_mode));
instance->fw_sync_cache_support = (scratch_pad_1 &
MR_CAN_HANDLE_SYNC_CACHE_OFFSET) ? 1 : 0;
@ -1190,9 +1190,17 @@ megasas_ioc_init_fusion(struct megasas_instance *instance)
* Each bit in replyqueue_mask represents one group of MSI-x vectors
* (each group has 8 vectors)
*/
if (instance->balanced_mode)
switch (instance->perf_mode) {
case MR_BALANCED_PERF_MODE:
init_frame->replyqueue_mask =
cpu_to_le16(~(~0 << instance->low_latency_index_start/8));
break;
case MR_IOPS_PERF_MODE:
init_frame->replyqueue_mask =
cpu_to_le16(~(~0 << instance->msix_vectors/8));
break;
}
req_desc.u.low = cpu_to_le32(lower_32_bits(cmd->frame_phys_addr));
req_desc.u.high = cpu_to_le32(upper_32_bits(cmd->frame_phys_addr));
@ -2831,7 +2839,7 @@ megasas_build_ldio_fusion(struct megasas_instance *instance,
fp_possible = (io_info.fpOkForIo > 0) ? true : false;
}
if (instance->balanced_mode &&
if ((instance->perf_mode == MR_BALANCED_PERF_MODE) &&
atomic_read(&scp->device->device_busy) >
(io_info.data_arms * MR_DEVICE_HIGH_IOPS_DEPTH))
cmd->request_desc->SCSIIO.MSIxIndex =
@ -3164,7 +3172,7 @@ megasas_build_syspd_fusion(struct megasas_instance *instance,
cmd->request_desc->SCSIIO.DevHandle = io_request->DevHandle;
if (instance->balanced_mode &&
if ((instance->perf_mode == MR_BALANCED_PERF_MODE) &&
atomic_read(&scmd->device->device_busy) > MR_DEVICE_HIGH_IOPS_DEPTH)
cmd->request_desc->SCSIIO.MSIxIndex =
mega_mod64((atomic64_add_return(1, &instance->high_iops_outstanding) /