Skip to contentSkip to conversation

nvme-pci: report the actual number of tagset maps

PATCH Applied Keith Busch 3 participants 3 messages Sep 26, 2022 – Sep 28, 2022 NVMe

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nvme-pci: report the actual number of tagset maps

From: Keith Busch <[email protected]>

We've been reporting 2 maps regardless of whether the module parameter asked for anything beyond the default queues. A consequence of this means that blk-mq will reinitialize the all the hardware contexts and io schedulers on every controller reset when the mapping is exactly the same as before. This unnecessary overhead is adding several milliseconds on a reset for environments that don't need it. Report the actual number of mappings in use.

---
 drivers/nvme/host/pci.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/nvme/host/pci.c +4 -2
index 70b1922c8953..59fa2490b646 100644
@@ -2526,9 +2526,11 @@ static void nvme_pci_alloc_tag_set(struct nvme_dev *dev)
 
 	set->ops = &nvme_mq_ops;
 	set->nr_hw_queues = dev->online_queues - 1;
-	set->nr_maps = 2; /* default + read */
+	set->nr_maps = 1;
+	if (dev->io_queues[HCTX_TYPE_READ])
+		set->nr_maps = 2;
 	if (dev->io_queues[HCTX_TYPE_POLL])
-		set->nr_maps++;
+		set->nr_maps = 3;
 	set->timeout = NVME_IO_TIMEOUT;
 	set->numa_node = dev->ctrl.numa_node;
 	set->queue_depth = min_t(unsigned, dev->q_depth, BLK_MQ_MAX_DEPTH) - 1;
Signature · 1 line
2.30.2