Skip to contentSkip to conversation

IB/rdmavt: add lock to call to rvt_error_qp to prevent a race condition

PATCH Applied Niels Dossche 3 participants 4 messages Feb 28, 2022 – Apr 8, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 IB/rdmavt: add lock to call to rvt_error_qp to prevent a race condition

The documentation of the function rvt_error_qp says both r_lock and s_lock need to be held when calling that function.
It also asserts using lockdep that both of those locks are held. However, the commit I referenced in Fixes accidentally makes the call to rvt_error_qp in rvt_ruc_loopback no longer covered by r_lock. This results in the lockdep assertion failing and also possibly in a race condition.

Fixes d757c60eca9b ("IB/rdmavt: Fix concurrency panics in QP post_send and modify to error")
---
 drivers/infiniband/sw/rdmavt/qp.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/infiniband/sw/rdmavt/qp.c +5 -1
index ae50b56e8913..8ef112f883a7 100644
@@ -3190,7 +3190,11 @@ void rvt_ruc_loopback(struct rvt_qp *sqp)
 	spin_lock_irqsave(&sqp->s_lock, flags);
 	rvt_send_complete(sqp, wqe, send_status);
 	if (sqp->ibqp.qp_type == IB_QPT_RC) {
-		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
+		int lastwqe;
+
+		spin_lock(&sqp->r_lock);
+		lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
+		spin_unlock(&sqp->r_lock);
 
 		sqp->s_flags &= ~RVT_S_BUSY;
 		spin_unlock_irqrestore(&sqp->s_lock, flags);
Signature · 1 line
2.35.1
maintainer
#1 lore raw
Quoted 8 lines · Niels Dossche

The documentation of the function rvt_error_qp says both r_lock and s_lock need to be held when calling that function.
It also asserts using lockdep that both of those locks are held. However, the commit I referenced in Fixes accidentally makes the call to rvt_error_qp in rvt_ruc_loopback no longer covered by r_lock. This results in the lockdep assertion failing and also possibly in a race condition.

Quoted 5 lines
Fixes d757c60eca9b ("IB/rdmavt: Fix concurrency panics in QP post_send and modify to error")
---
 drivers/infiniband/sw/rdmavt/qp.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Dennis?

And this too:

https://patchwork.kernel.org/project/linux-rdma/patch/[email protected]/

Thanks,
Jason

maintainer Acked-by Acked-by
#2 lore raw
Quoted 9 lines · Jason Gunthorpe
Quoted 8 lines · Niels Dossche

The documentation of the function rvt_error_qp says both r_lock and s_lock need to be held when calling that function.
It also asserts using lockdep that both of those locks are held. However, the commit I referenced in Fixes accidentally makes the call to rvt_error_qp in rvt_ruc_loopback no longer covered by r_lock. This results in the lockdep assertion failing and also possibly in a race condition.

Quoted 7 lines
Quoted 5 lines
Fixes d757c60eca9b ("IB/rdmavt: Fix concurrency panics in QP post_send and modify to error")
---
 drivers/infiniband/sw/rdmavt/qp.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Dennis?

For this one:

Quoted 3 lines

And this too:

https://patchwork.kernel.org/project/linux-rdma/patch/[email protected]/

I think is correct as well. Not thrilled with the lock, unlock, lock. Stuff, but this whole unwind code needs a once-over so for this patch as well:

Thanks Niels for finding and fixing these issues!

-Denny

maintainer Applied
#3 lore raw
In reply to Niels Dossche
Quoted 8 lines · Niels Dossche

The documentation of the function rvt_error_qp says both r_lock and s_lock need to be held when calling that function.
It also asserts using lockdep that both of those locks are held. However, the commit I referenced in Fixes accidentally makes the call to rvt_error_qp in rvt_ruc_loopback no longer covered by r_lock. This results in the lockdep assertion failing and also possibly in a race condition.

Quoted 5 lines
Fixes d757c60eca9b ("IB/rdmavt: Fix concurrency panics in QP post_send and modify to error")
---
 drivers/infiniband/sw/rdmavt/qp.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Applied to for-rc, thanks

Jason