Skip to contentSkip to conversation

nvmet-rdma: fix null dereference under heavy load

PATCH Superseded Raju Rangoju 3 participants 4 messages Jan 3, 2019

Subsystems (3)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 nvmet-rdma: fix null dereference under heavy load

Under heavy load if we don't have any pre-allocated rsps left, we dynamically allocate a rsp, but we are not actually allocating memory for nvme_completion (rsp->req.rsp). In such a case, accessing pointer fields (req->rsp->status) in nvmet_req_init() will result in crash.

To fix this, allocate the memory for nvme_completion by calling nvmet_rdma_alloc_rsp()

Fixes 8407879c("nvmet-rdma:fix possible bogus dereference under heavy load")
Cc <stable at vger.kernel.org>
Reviewed-by Max Gurtovoy <maxg at mellanox.com>
Signed-off-by Raju Rangoju <rajur at chelsio.com>

--
Changes from v1:
- Moved integer to 'if' block
- Used 'unlikely' in datapath flow condition

---
 drivers/nvme/target/rdma.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)
drivers/nvme/target/rdma.c +14 -1
index a8d23eb80192..8f9e6645fcf9 100644
@@ -139,6 +139,10 @@ static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc);
 static void nvmet_rdma_read_data_done(struct ib_cq *cq, struct ib_wc *wc);
 static void nvmet_rdma_qp_event(struct ib_event *event, void *priv);
 static void nvmet_rdma_queue_disconnect(struct nvmet_rdma_queue *queue);
+static void nvmet_rdma_free_rsp(struct nvmet_rdma_device *ndev,
+				struct nvmet_rdma_rsp *r);
+static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev,
+				struct nvmet_rdma_rsp *r);
 
 static const struct nvmet_fabrics_ops nvmet_rdma_ops;
 
@@ -182,9 +186,17 @@ nvmet_rdma_get_rsp(struct nvmet_rdma_queue *queue)
 	spin_unlock_irqrestore(&queue->rsps_lock, flags);
 
 	if (unlikely(!rsp)) {
-		rsp = kmalloc(sizeof(*rsp), GFP_KERNEL);
+		int ret = -EINVAL;
+
+		rsp = kzalloc(sizeof(*rsp), GFP_KERNEL);
 		if (unlikely(!rsp))
 			return NULL;
+		ret = nvmet_rdma_alloc_rsp(queue->dev, rsp);
+		if (unlikely(ret)) {
+			kfree(rsp);
+			return NULL;
+		}
+
 		rsp->allocated = true;
 	}
 
@@ -197,6 +209,7 @@ nvmet_rdma_put_rsp(struct nvmet_rdma_rsp *rsp)
 	unsigned long flags;
 
 	if (unlikely(rsp->allocated)) {
+		nvmet_rdma_free_rsp(rsp->queue->dev, rsp);
 		kfree(rsp);
 		return;
 	}
Signature · 1 line
2.12.0
#1 lore raw

PATCH v2 1/1 nvmet-rdma: fix null dereference under heavy load

Quoted 38 lines · Raju Rangoju

Under heavy load if we don't have any pre-allocated rsps left, we dynamically allocate a rsp, but we are not actually allocating memory for nvme_completion (rsp->req.rsp). In such a case, accessing pointer fields (req->rsp->status) in nvmet_req_init() will result in crash.

To fix this, allocate the memory for nvme_completion by calling nvmet_rdma_alloc_rsp()

Fixes 8407879c("nvmet-rdma:fix possible bogus dereference under heavy load")
Cc <stable at vger.kernel.org>
Reviewed-by Max Gurtovoy <maxg at mellanox.com>
Signed-off-by Raju Rangoju <rajur at chelsio.com>

--
Changes from v1:
- Moved integer to 'if' block
- Used 'unlikely' in datapath flow condition

---
  drivers/nvme/target/rdma.c | 15 ++++++++++++++-
  1 file changed, 14 insertions(+), 1 deletion(-)
drivers/nvme/target/rdma.c +4 -0
index a8d23eb80192..8f9e6645fcf9 100644
@@ -139,6 +139,10 @@ static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc);
  static void nvmet_rdma_read_data_done(struct ib_cq *cq, struct ib_wc *wc);
  static void nvmet_rdma_qp_event(struct ib_event *event, void *priv);
  static void nvmet_rdma_queue_disconnect(struct nvmet_rdma_queue *queue);
+static void nvmet_rdma_free_rsp(struct nvmet_rdma_device *ndev,
+				struct nvmet_rdma_rsp *r);
+static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev,
+				struct nvmet_rdma_rsp *r);
 
  static const struct nvmet_fabrics_ops nvmet_rdma_ops;
 

drivers/nvme/target/rdma.c nvmet_rdma_get_rsp(struct nvmet_rdma_queue *queue) · line 190

Quoted 6 lines
drivers/nvme/target/rdma.c +2 -1
@@ -182,9 +186,17 @@ nvmet_rdma_get_rsp(struct nvmet_rdma_queue *queue)
  	spin_unlock_irqrestore(&queue->rsps_lock, flags);
 
  	if (unlikely(!rsp)) {
-		rsp = kmalloc(sizeof(*rsp), GFP_KERNEL);
+		int ret = -EINVAL;
+

no real need to initialize ret variable (sorry I didn't see it in first review).

Quoted 20 lines
drivers/nvme/target/rdma.c +8 -0
+		rsp = kzalloc(sizeof(*rsp), GFP_KERNEL);
  		if (unlikely(!rsp))
  			return NULL;
+		ret = nvmet_rdma_alloc_rsp(queue->dev, rsp);
+		if (unlikely(ret)) {
+			kfree(rsp);
+			return NULL;
+		}
+
  		rsp->allocated = true;
  	}
 
@@ -197,6 +209,7 @@ nvmet_rdma_put_rsp(struct nvmet_rdma_rsp *rsp)
  	unsigned long flags;
 
  	if (unlikely(rsp->allocated)) {
+		nvmet_rdma_free_rsp(rsp->queue->dev, rsp);
  		kfree(rsp);
  		return;
  	}
#2 lore raw

PATCH v2 1/1 nvmet-rdma: fix null dereference under heavy load

Quoted 11 lines · Max Gurtovoy

drivers/nvme/target/rdma.c nvmet_rdma_get_rsp(struct nvmet_rdma_queue *queue) · line 190

Quoted 6 lines
drivers/nvme/target/rdma.c +2 -1
 	spin_unlock_irqrestore(&queue->rsps_lock, flags);
 	if (unlikely(!rsp)) {
-		rsp = kmalloc(sizeof(*rsp), GFP_KERNEL);
+		int ret = -EINVAL;
+

no real need to initialize ret variable (sorry I didn't see it in first review).

No problem. I'll post v3.

Quoted 1 line
Quoted 1 line
drivers/nvme/target/rdma.c +1 -0
+		rsp = kzalloc(sizeof(*rsp), GFP_KERNEL);
maintainer
#3 lore raw

PATCH v2 1/1 nvmet-rdma: fix null dereference under heavy load

In reply to Raju Rangoju

Hi,

[This is an automated email]

This commit has been processed because it contains a "Fixes:" tag, fixing commit: .

The bot has tested the following trees: v4.20.0, v4.19.13, v4.14.91, v4.9.148, v4.4.169, v3.18.131,

v4.20.0: Failed to apply! Possible dependencies:
ad1f824948e4 ("nvmet-rdma: Add unlikely for response allocated check")

v4.19.13: Failed to apply! Possible dependencies:
ad1f824948e4 ("nvmet-rdma: Add unlikely for response allocated check")

v4.14.91: Failed to apply! Possible dependencies:
ad1f824948e4 ("nvmet-rdma: Add unlikely for response allocated check")

v4.9.148: Failed to apply! Possible dependencies:
ad1f824948e4 ("nvmet-rdma: Add unlikely for response allocated check")

v4.4.169: Failed to apply! Possible dependencies:
3a85a5de29ea ("nvme-loop: add a NVMe loopback host driver") 748ff8408f8e ("nvmet-rdma: Fix missing dma sync to nvme data structures") 8407879c4e0d ("nvmet-rdma: fix possible bogus dereference under heavy load") 8d61413db604 ("nvmet-rdma: Fix a possible uninitialized variable dereference") 8f000cac6e7a ("nvmet-rdma: add a NVMe over Fabrics RDMA target driver") a07b4970f464 ("nvmet: add a generic NVMe target")
ad1f824948e4 ("nvmet-rdma: Add unlikely for response allocated check")

v3.18.131: Failed to apply! Possible dependencies:
062261be4e39 ("nvme: Replace rcu_assign_pointer() with RCU_INIT_POINTER()") 1d0906246095 ("NVMe: Mismatched host/device page size support") 302c6727e5eb ("NVMe: Fix filesystem sync deadlock on removal") 57dacad5f228 ("nvme: move to a new drivers/nvme/host directory") 5905535610fc ("NVMe: Correctly handle IOCTL_SUBMIT_IO when cpus > online queues") 5940c8578fe7 ("NVMe: Clear QUEUE_FLAG_STACKABLE")
6fccf9383b28 ("NVMe: Async event request")
7963e521811e ("NVMe: Passthrough IOCTL for IO commands") 7be50e93fbc2 ("NVMe: Fix nvmeq waitqueue entry initialization") 8f000cac6e7a ("nvmet-rdma: add a NVMe over Fabrics RDMA target driver") 9dbbfab7d541 ("NVMe: Do not over allocate for discard requests") a07b4970f464 ("nvmet: add a generic NVMe target")
a4aea5623d4a ("NVMe: Convert to blk-mq")
a96d4f5c2da6 ("NVMe: Reference count pci device")
aee4b9bd4505 ("nvmem: Add to MAINTAINERS for nvmem framework") b4ff9c8ddb6f ("NVMe: Translate NVMe status to errno")
dece45855a8b ("NFC: nxp-nci: Add support for NXP NCI chips") f435c2825b4c ("NVMe: Call nvme_free_queue directly")

How should we proceed with this patch?

--
Thanks,
Sasha