Skip to contentSkip to conversation

nvme: fine-tune sending of first keep-alive

PATCH RESEND Applied Mark O'Donovan 3 participants 3 messages Nov 24, 2023 – Nov 27, 2023 NVMe

lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 nvme: fine-tune sending of first keep-alive

Keep-alive commands are sent half-way through the kato period. This normally works well but fails when the keep-alive system is started when we are more than half way through the kato. This can happen on larger setups or due to host delays. With this change we now time the initial keep-alive command from the controller initialisation time, rather than the keep-alive mechanism activation time.

---
 drivers/nvme/host/core.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
drivers/nvme/host/core.c +11 -2
index 46a4c9c5ea96..8bf24c1cd8bb 100644
@@ -1192,8 +1192,16 @@ static unsigned long nvme_keep_alive_work_period(struct nvme_ctrl *ctrl)
 
 static void nvme_queue_keep_alive_work(struct nvme_ctrl *ctrl)
 {
-	queue_delayed_work(nvme_wq, &ctrl->ka_work,
-			   nvme_keep_alive_work_period(ctrl));
+	unsigned long now = jiffies;
+	unsigned long delay = nvme_keep_alive_work_period(ctrl);
+	unsigned long ka_next_check_tm = ctrl->ka_last_check_time + delay;
+
+	if (time_after(now, ka_next_check_tm))
+		delay = 0;
+	else
+		delay = ka_next_check_tm - now;
+
+	queue_delayed_work(nvme_wq, &ctrl->ka_work, delay);
 }
 
 static enum rq_end_io_ret nvme_keep_alive_end_io(struct request *rq,
@@ -4471,6 +4479,7 @@ int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev,
 	INIT_DELAYED_WORK(&ctrl->failfast_work, nvme_failfast_work);
 	memset(&ctrl->ka_cmd, 0, sizeof(ctrl->ka_cmd));
 	ctrl->ka_cmd.common.opcode = nvme_admin_keep_alive;
+	ctrl->ka_last_check_time = jiffies;
 
 	BUILD_BUG_ON(NVME_DSM_MAX_RANGES * sizeof(struct nvme_dsm_range) >
 			PAGE_SIZE);

base-commit: 5b7ad877e4d81f8904ce83982b1ba5c6e83deccb

Signature · 1 line
2.39.2
maintainer Applied
#2 lore raw
In reply to Mark O'Donovan
Quoted 7 lines · Mark O'Donovan

Keep-alive commands are sent half-way through the kato period. This normally works well but fails when the keep-alive system is started when we are more than half way through the kato. This can happen on larger setups or due to host delays. With this change we now time the initial keep-alive command from the controller initialisation time, rather than the keep-alive mechanism activation time.

Thanks, applied to nvme-6.7.