An approximation for the PacketLifeTime is half the local ACK timeout.
The encoding for both timers are logarithmic. The PacketLifeTime
calculation is wrong when local ACK timeout is zero. In that case,
PacketLifeTime is set to the incorrect value 255.
Fixed by explicitly testing for timeout being zero.
Fixes: e1ee1e62bec4 ("RDMA/cma: Use ACK timeout for RoCE packetLifeTime")
Signed-off-by: Håkon Bugge <haakon.bugge@xxxxxxxxxx>
---
* Note: This commit must be merged after ("RDMA/cma: Replace
RMW with atomic bit-ops")
---
drivers/infiniband/core/cma.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index 6759889..b1512ca 100644
--- a/drivers/infiniband/core/cma.c
+++ b/drivers/infiniband/core/cma.c
@@ -3096,9 +3096,11 @@ static int cma_resolve_iboe_route(struct rdma_id_private *id_priv)
* PacketLifeTime = local ACK timeout/2
* as a reasonable approximation for RoCE networks.
*/
- route->path_rec->packet_life_time =
- test_bit(TIMEOUT_SET, &id_priv->flags) ?
- id_priv->timeout - 1 : CMA_IBOE_PACKET_LIFETIME;
+ if (test_bit(TIMEOUT_SET, &id_priv->flags))
+ route->path_rec->packet_life_time =
+ id_priv->timeout ? id_priv->timeout - 1 : 0;
+ else
+ route->path_rec->packet_life_time = CMA_IBOE_PACKET_LIFETIME;
if (!route->path_rec->mtu) {
ret = -EINVAL;
--
1.8.3.1
