Commit addb8a65 authored by Jack Morgenstein's avatar Jack Morgenstein Committed by Jason Gunthorpe

RDMA/uverbs: Expand primary and alt AV port checks

The commit cited below checked that the port numbers provided in the
primary and alt AVs are legal.

That is sufficient to prevent a kernel panic. However, it is not
sufficient for correct operation.

In Linux, AVs (both primary and alt) must be completely self-described.
We do not accept an AV from userspace without an embedded port number.
(This has been the case since kernel 3.14 commit dbf727de
("IB/core: Use GID table in AH creation and dmac resolution")).

For the primary AV, this embedded port number must match the port number
specified with IB_QP_PORT.

We also expect the port number embedded in the alt AV to match the
alt_port_num value passed by the userspace driver in the modify_qp command
base structure.

Add these checks to modify_qp.

Cc: <stable@vger.kernel.org> # 4.16
Fixes: 5d4c05c3 ("RDMA/uverbs: Sanitize user entered port numbers prior to access it")
Signed-off-by: default avatarJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: default avatarLeon Romanovsky <leonro@mellanox.com>
Signed-off-by: default avatarJason Gunthorpe <jgg@mellanox.com>
parent d63c4673
...@@ -1984,15 +1984,64 @@ static int modify_qp(struct ib_uverbs_file *file, ...@@ -1984,15 +1984,64 @@ static int modify_qp(struct ib_uverbs_file *file,
goto release_qp; goto release_qp;
} }
if ((cmd->base.attr_mask & IB_QP_AV) && if ((cmd->base.attr_mask & IB_QP_AV)) {
!rdma_is_port_valid(qp->device, cmd->base.dest.port_num)) { if (!rdma_is_port_valid(qp->device, cmd->base.dest.port_num)) {
ret = -EINVAL; ret = -EINVAL;
goto release_qp; goto release_qp;
}
if (cmd->base.attr_mask & IB_QP_STATE &&
cmd->base.qp_state == IB_QPS_RTR) {
/* We are in INIT->RTR TRANSITION (if we are not,
* this transition will be rejected in subsequent checks).
* In the INIT->RTR transition, we cannot have IB_QP_PORT set,
* but the IB_QP_STATE flag is required.
*
* Since kernel 3.14 (commit dbf727de7440), the uverbs driver,
* when IB_QP_AV is set, has required inclusion of a valid
* port number in the primary AV. (AVs are created and handled
* differently for infiniband and ethernet (RoCE) ports).
*
* Check the port number included in the primary AV against
* the port number in the qp struct, which was set (and saved)
* in the RST->INIT transition.
*/
if (cmd->base.dest.port_num != qp->real_qp->port) {
ret = -EINVAL;
goto release_qp;
}
} else {
/* We are in SQD->SQD. (If we are not, this transition will
* be rejected later in the verbs layer checks).
* Check for both IB_QP_PORT and IB_QP_AV, these can be set
* together in the SQD->SQD transition.
*
* If only IP_QP_AV was set, add in IB_QP_PORT as well (the
* verbs layer driver does not track primary port changes
* resulting from path migration. Thus, in SQD, if the primary
* AV is modified, the primary port should also be modified).
*
* Note that in this transition, the IB_QP_STATE flag
* is not allowed.
*/
if (((cmd->base.attr_mask & (IB_QP_AV | IB_QP_PORT))
== (IB_QP_AV | IB_QP_PORT)) &&
cmd->base.port_num != cmd->base.dest.port_num) {
ret = -EINVAL;
goto release_qp;
}
if ((cmd->base.attr_mask & (IB_QP_AV | IB_QP_PORT))
== IB_QP_AV) {
cmd->base.attr_mask |= IB_QP_PORT;
cmd->base.port_num = cmd->base.dest.port_num;
}
}
} }
if ((cmd->base.attr_mask & IB_QP_ALT_PATH) && if ((cmd->base.attr_mask & IB_QP_ALT_PATH) &&
(!rdma_is_port_valid(qp->device, cmd->base.alt_port_num) || (!rdma_is_port_valid(qp->device, cmd->base.alt_port_num) ||
!rdma_is_port_valid(qp->device, cmd->base.alt_dest.port_num))) { !rdma_is_port_valid(qp->device, cmd->base.alt_dest.port_num) ||
cmd->base.alt_port_num != cmd->base.alt_dest.port_num)) {
ret = -EINVAL; ret = -EINVAL;
goto release_qp; goto release_qp;
} }
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment