Skip to contentSkip to conversation

IB/{hfi1,qib}: Add limit test for RC/UC send via loopback

PATCH In review Mike Marciniszyn 3 participants 4 messages Feb 4, 2019 – Feb 11, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 IB/{hfi1,qib}: Add limit test for RC/UC send via loopback

Upstream commit 09ce351dff8e7636af0beb72cd4a86c3904a0500.

Fix potential memory corruption and panic in loopback for IB_WR_SEND varaiants.

The code blindly assumes the posted length will fit in the fetched rwqe, which is not a valid assumption.

Fix by adding a limit test, and triggering the appropriate send completion and putting the QP in an error state. This mimics the handling for non-loopback QPs.

[ported to older kernels where the loopback was in each driver]

Cc <[email protected]> #v4.9+
Fixes 7724105686e7 ("IB/hfi1: add driver files")
---
 drivers/infiniband/hw/hfi1/ruc.c    |    7 ++++++-
 drivers/infiniband/hw/qib/qib_ruc.c |    7 ++++++-
 2 files changed, 12 insertions(+), 2 deletions(-)
drivers/infiniband/hw/hfi1/ruc.c +6 -1
index 5f56f3c..62a3832 100644
@@ -278,6 +278,8 @@ static void ruc_loopback(struct rvt_qp *sqp)
 			goto op_err;
 		if (!ret)
 			goto rnr_nak;
+		if (wqe->length > qp->r_len)
+			goto inv_err;
 		break;
 
 	case IB_WR_RDMA_WRITE_WITH_IMM:
@@ -445,7 +447,10 @@ static void ruc_loopback(struct rvt_qp *sqp)
 	goto err;
 
 inv_err:
-	send_status = IB_WC_REM_INV_REQ_ERR;
+	send_status =
+		sqp->ibqp.qp_type == IB_QPT_RC ?
+			IB_WC_REM_INV_REQ_ERR :
+			IB_WC_SUCCESS;
 	wc.status = IB_WC_LOC_QP_OP_ERR;
 	goto err;
 
drivers/infiniband/hw/qib/qib_ruc.c +6 -1
index f8a7de7..563f71e 100644
@@ -274,6 +274,8 @@ static void qib_ruc_loopback(struct rvt_qp *sqp)
 			goto op_err;
 		if (!ret)
 			goto rnr_nak;
+		if (wqe->length > qp->r_len)
+			goto inv_err;
 		break;
 
 	case IB_WR_RDMA_WRITE_WITH_IMM:
@@ -434,7 +436,10 @@ static void qib_ruc_loopback(struct rvt_qp *sqp)
 	goto err;
 
 inv_err:
-	send_status = IB_WC_REM_INV_REQ_ERR;
+	send_status =
+		sqp->ibqp.qp_type == IB_QPT_RC ?
+			IB_WC_REM_INV_REQ_ERR :
+			IB_WC_SUCCESS;
 	wc.status = IB_WC_LOC_QP_OP_ERR;
 	goto err;
 
maintainer
#1 lore raw
Quoted 14 lines · Mike Marciniszyn

Upstream commit 09ce351dff8e7636af0beb72cd4a86c3904a0500.

Fix potential memory corruption and panic in loopback for IB_WR_SEND varaiants.

The code blindly assumes the posted length will fit in the fetched rwqe, which is not a valid assumption.

Fix by adding a limit test, and triggering the appropriate send completion and putting the QP in an error state. This mimics the handling for non-loopback QPs.

[ported to older kernels where the loopback was in each driver]

Quoted 8 lines
Cc <[email protected]> #v4.9+
Fixes 7724105686e7 ("IB/hfi1: add driver files")
---
 drivers/infiniband/hw/hfi1/ruc.c    |    7 ++++++-
 drivers/infiniband/hw/qib/qib_ruc.c |    7 ++++++-
 2 files changed, 12 insertions(+), 2 deletions(-)

What stable kernel(s) is this backport for?

thanks,

greg k-h

#2 lore raw
Quoted 7 lines

Subject: Re: [PATCH] IB/{hfi1,qib}: Add limit test for RC/UC send via loopback

What stable kernel(s) is this backport for?

thanks,

greg k-h

The original patch has already been backported to 4.20.

The kernels that require the ported patch are older ones: 4.19.x, 4.14.x, and 4.9.x.

Mike

#3 lore raw
In reply to Mike Marciniszyn
Quoted 14 lines · Mike Marciniszyn

Upstream commit 09ce351dff8e7636af0beb72cd4a86c3904a0500.

Fix potential memory corruption and panic in loopback for IB_WR_SEND varaiants.

The code blindly assumes the posted length will fit in the fetched rwqe, which is not a valid assumption.

Fix by adding a limit test, and triggering the appropriate send completion and putting the QP in an error state. This mimics the handling for non-loopback QPs.

[ported to older kernels where the loopback was in each driver]

Quoted 8 lines
Cc <[email protected]> #v4.9+
Fixes 7724105686e7 ("IB/hfi1: add driver files")
---
 drivers/infiniband/hw/hfi1/ruc.c    |    7 ++++++-
 drivers/infiniband/hw/qib/qib_ruc.c |    7 ++++++-
 2 files changed, 12 insertions(+), 2 deletions(-)

Now queued up, thanks.

greg k-h