| Message ID | 1624371207-26710-1-git-send-email-haakon.bugge@oracle.com (mailing list archive) |
|---|---|
| State | Accepted |
| Delegated to: | Jason Gunthorpe |
| Headers | show |
| Series | [for-next,v2] RDMA/cma: Fix incorrect Packet Lifetime calculation | expand |
On Tue, Jun 22, 2021 at 04:13:27PM +0200, Håkon Bugge wrote: > An approximation for the PacketLifeTime is half the local ACK timeout. > The encoding for both timers are logarithmic. > > If the local ACK timeout is set, but zero, it means the timer is > disabled. In this case, we choose the CMA_IBOE_PACKET_LIFETIME value, > since 50% of infinite makes no sense. > > Before this commit, the PacketLifeTime became 255 if local ACK > timeout was zero (not running). > > 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@oracle.com> > Reviewed-by: Leon Romanovsky <leonro@nvidia.com> > --- > > v1 -> v2: > * Rebased on tip of for-next with ("RDMA/cma: Protect RMW > with qp_mutex") included > * A local ACK timeout of zero now sets PacketLifeTime to > CMA_IBOE_PACKET_LIFETIME > * Added Leon's r-b > --- > drivers/infiniband/core/cma.c | 6 ++++-- > 1 file changed, 4 insertions(+), 2 deletions(-) Applied to for-next, thanks Jason
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index c44a0c4..35fe300 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -3080,8 +3080,10 @@ static int cma_resolve_iboe_route(struct rdma_id_private *id_priv) * as a reasonable approximation for RoCE networks. */ mutex_lock(&id_priv->qp_mutex); - route->path_rec->packet_life_time = id_priv->timeout_set ? - id_priv->timeout - 1 : CMA_IBOE_PACKET_LIFETIME; + if (id_priv->timeout_set && id_priv->timeout) + route->path_rec->packet_life_time = id_priv->timeout - 1; + else + route->path_rec->packet_life_time = CMA_IBOE_PACKET_LIFETIME; mutex_unlock(&id_priv->qp_mutex); if (!route->path_rec->mtu) {
An approximation for the PacketLifeTime is half the local ACK timeout. The encoding for both timers are logarithmic. If the local ACK timeout is set, but zero, it means the timer is disabled. In this case, we choose the CMA_IBOE_PACKET_LIFETIME value, since 50% of infinite makes no sense. Before this commit, the PacketLifeTime became 255 if local ACK timeout was zero (not running). 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@oracle.com> Reviewed-by: Leon Romanovsky <leonro@nvidia.com> --- v1 -> v2: * Rebased on tip of for-next with ("RDMA/cma: Protect RMW with qp_mutex") included * A local ACK timeout of zero now sets PacketLifeTime to CMA_IBOE_PACKET_LIFETIME * Added Leon's r-b --- drivers/infiniband/core/cma.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)