Skip to contentSkip to conversation

nvme: free sq/cq dbbuf pointers when dbbuf set fails

PATCH Applied Minwoo Im 3 participants 4 messages Nov 5, 2020 – Nov 14, 2020 NVMe

lore
Share & tools
lore

Minwoo Im

author
#0 lore raw

PATCH 1/1 nvme: free sq/cq dbbuf pointers when dbbuf set fails

If Doorbell Buffer Config command fails even 'dev->dbbuf_dbs != NULL' which means OACS indicates that NVME_CTRL_OACS_DBBUF_SUPP is set, nvme_dbbuf_update_and_check_event() will check event even it's not been successfully set.

This patch fixes mismatch among dbbuf for sq/cqs in case that dbbuf command fails.

---
 drivers/nvme/host/pci.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
drivers/nvme/host/pci.c +15 -0
index 0578ff253c47..3be352403839 100644
@@ -292,9 +292,21 @@ static void nvme_dbbuf_init(struct nvme_dev *dev,
 	nvmeq->dbbuf_cq_ei = &dev->dbbuf_eis[cq_idx(qid, dev->db_stride)];
 }
 
+static void nvme_dbbuf_free(struct nvme_queue *nvmeq)
+{
+	if (!nvmeq->qid)
+		return;
+
+	nvmeq->dbbuf_sq_db = NULL;
+	nvmeq->dbbuf_cq_db = NULL;
+	nvmeq->dbbuf_sq_ei = NULL;
+	nvmeq->dbbuf_cq_ei = NULL;
+}
+
 static void nvme_dbbuf_set(struct nvme_dev *dev)
 {
 	struct nvme_command c;
+	unsigned int i;
 
 	if (!dev->dbbuf_dbs)
 		return;
@@ -308,6 +320,9 @@ static void nvme_dbbuf_set(struct nvme_dev *dev)
 		dev_warn(dev->ctrl.device, "unable to set dbbuf\n");
 		/* Free memory and continue on */
 		nvme_dbbuf_dma_free(dev);
+
+		for (i = 1; i <= dev->online_queues; i++)
+			nvme_dbbuf_free(&dev->queues[i]);
 	}
 }
 
Signature · 7 lines
2.17.1


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
maintainer
#1 lore raw
Quoted 7 lines · Minwoo Im

If Doorbell Buffer Config command fails even 'dev->dbbuf_dbs != NULL' which means OACS indicates that NVME_CTRL_OACS_DBBUF_SUPP is set, nvme_dbbuf_update_and_check_event() will check event even it's not been successfully set.

This patch fixes mismatch among dbbuf for sq/cqs in case that dbbuf command fails.

This fixes if host dbbuf was successfully configured at one point, but fails on a subsequent reset, correct? Otherwise, I don't see how the queue's dbbuf values would have been set to non-NULL.

_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme

author
#2 lore raw
Quoted 7 lines · Keith Busch
Quoted 5 lines · Minwoo Im

If Doorbell Buffer Config command fails even 'dev->dbbuf_dbs != NULL' which means OACS indicates that NVME_CTRL_OACS_DBBUF_SUPP is set, nvme_dbbuf_update_and_check_event() will check event even it's not been successfully set.

Quoted 6 lines
Quoted 2 lines

This patch fixes mismatch among dbbuf for sq/cqs in case that dbbuf command fails.

This fixes if host dbbuf was successfully configured at one point, but fails on a subsequent reset, correct? Otherwise, I don't see how the queue's dbbuf values would have been set to non-NULL.

Yes. reset_work() checks OACS with DBBUF_SUPP and if it's reported as supported, then it will alloc dev->dbbuf_dbs. After that, However, if nvme_dbbuf_set() fails with sc != 0 then it should free for each queue's dbbuf pointer to prevent being checked for events when updating the doorbell.

_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme