| Message ID | 20220228195144.71946-1-dossche.niels@gmail.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 22cbc6c2681a0a4fe76150270426e763d52353a4 |
| Headers | show |
| Series | IB/rdmavt: add missing locks in rvt_ruc_loopback | expand |
On Mon, Feb 28, 2022 at 08:51:44PM +0100, Niels Dossche wrote: > 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. > rvt_error_qp is called form rvt_send_cq, which is called from > rvt_qp_complete_swqe, which is called from rvt_send_complete, which is > called from rvt_ruc_loopback in two places. Both of these places do not > hold r_lock. Fix this by acquiring a spin_lock of r_lock in both of > these places. > The r_lock acquiring cannot be added in rvt_qp_complete_swqe because > some of its other callers already have r_lock acquired. > > Signed-off-by: Niels Dossche <dossche.niels@gmail.com> > --- > drivers/infiniband/sw/rdmavt/qp.c | 6 +++++- > 1 file changed, 5 insertions(+), 1 deletion(-) Applied to for-next, thanks Jason
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c index ae50b56e8913..3bac4f90a6d6 100644 --- a/drivers/infiniband/sw/rdmavt/qp.c +++ b/drivers/infiniband/sw/rdmavt/qp.c @@ -2775,7 +2775,7 @@ void rvt_qp_iter(struct rvt_dev_info *rdi, EXPORT_SYMBOL(rvt_qp_iter); /* - * This should be called with s_lock held. + * This should be called with s_lock and r_lock held. */ void rvt_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe, enum ib_wc_status status) @@ -3134,7 +3134,9 @@ void rvt_ruc_loopback(struct rvt_qp *sqp) rvp->n_loop_pkts++; flush_send: sqp->s_rnr_retry = sqp->s_rnr_retry_cnt; + spin_lock(&sqp->r_lock); rvt_send_complete(sqp, wqe, send_status); + spin_unlock(&sqp->r_lock); if (local_ops) { atomic_dec(&sqp->local_ops_pending); local_ops = 0; @@ -3188,7 +3190,9 @@ void rvt_ruc_loopback(struct rvt_qp *sqp) spin_unlock_irqrestore(&qp->r_lock, flags); serr_no_r_lock: spin_lock_irqsave(&sqp->s_lock, flags); + spin_lock(&sqp->r_lock); rvt_send_complete(sqp, wqe, send_status); + spin_unlock(&sqp->r_lock); if (sqp->ibqp.qp_type == IB_QPT_RC) { int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
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. rvt_error_qp is called form rvt_send_cq, which is called from rvt_qp_complete_swqe, which is called from rvt_send_complete, which is called from rvt_ruc_loopback in two places. Both of these places do not hold r_lock. Fix this by acquiring a spin_lock of r_lock in both of these places. The r_lock acquiring cannot be added in rvt_qp_complete_swqe because some of its other callers already have r_lock acquired. Signed-off-by: Niels Dossche <dossche.niels@gmail.com> --- drivers/infiniband/sw/rdmavt/qp.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)