Skip to contentSkip to conversation

RDMA/usnic: Remove the query_pkey callback

PATCH In review Kamal Heib 4 participants 12 messages Aug 20, 2020 – Aug 27, 2020

Subsystems (2)
lore
Share & tools
lore

Kamal Heib

author
#0 lore raw

PATCH 1/1 RDMA/usnic: Remove the query_pkey callback

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

---
 drivers/infiniband/hw/usnic/usnic_ib_main.c  |  2 --
 drivers/infiniband/hw/usnic/usnic_ib_verbs.c | 11 -----------
 drivers/infiniband/hw/usnic/usnic_ib_verbs.h |  2 --
 3 files changed, 15 deletions(-)
drivers/infiniband/hw/usnic/usnic_ib_main.c +0 -2
index c9abe1c01e4e..515460c4212d 100644
@@ -315,7 +315,6 @@ static int usnic_port_immutable(struct ib_device *ibdev, u8 port_num,
 	if (err)
 		return err;
 
-	immutable->pkey_tbl_len = attr.pkey_tbl_len;
 	immutable->gid_tbl_len = attr.gid_tbl_len;
 
 	return 0;
@@ -355,7 +354,6 @@ static const struct ib_device_ops usnic_dev_ops = {
 	.modify_qp = usnic_ib_modify_qp,
 	.query_device = usnic_ib_query_device,
 	.query_gid = usnic_ib_query_gid,
-	.query_pkey = usnic_ib_query_pkey,
 	.query_port = usnic_ib_query_port,
 	.query_qp = usnic_ib_query_qp,
 	.reg_user_mr = usnic_ib_reg_mr,
drivers/infiniband/hw/usnic/usnic_ib_verbs.c +0 -11
index b8a77ce11590..02a49f661c8d 100644
@@ -367,7 +367,6 @@ int usnic_ib_query_port(struct ib_device *ibdev, u8 port,
 
 	props->port_cap_flags = 0;
 	props->gid_tbl_len = 1;
-	props->pkey_tbl_len = 1;
 	props->bad_pkey_cntr = 0;
 	props->qkey_viol_cntr = 0;
 	props->max_mtu = IB_MTU_4096;
@@ -437,16 +436,6 @@ int usnic_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
 	return 0;
 }
 
-int usnic_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
-				u16 *pkey)
-{
-	if (index > 0)
-		return -EINVAL;
-
-	*pkey = 0xffff;
-	return 0;
-}
-
 int usnic_ib_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
 {
 	struct usnic_ib_pd *pd = to_upd(ibpd);
drivers/infiniband/hw/usnic/usnic_ib_verbs.h +0 -2
index 2aedf78c13cf..9195f2b901ce 100644
@@ -48,8 +48,6 @@ int usnic_ib_query_qp(struct ib_qp *qp, struct ib_qp_attr *qp_attr,
 				struct ib_qp_init_attr *qp_init_attr);
 int usnic_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
 				union ib_gid *gid);
-int usnic_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
-				u16 *pkey);
 int usnic_ib_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata);
 void usnic_ib_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata);
 struct ib_qp *usnic_ib_create_qp(struct ib_pd *pd,
Signature · 1 line
2.26.2
#1 lore raw
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

author
#2 lore raw
Quoted 6 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Or changed to work only for node_type == e.IBV_NODE_CA?

Thanks,
Kamal

author
#3 lore raw
Quoted 5 lines · Kamal Heib
Quoted 4 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Quoted 7 lines
Quoted 2 lines

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Or changed to work only for node_type == e.IBV_NODE_CA?

Thanks,
Kamal

BTW, do the efa care about pkey?

Thanks,
Kamal

#4 lore raw
Quoted 9 lines · Kamal Heib
Quoted 8 lines · Kamal Heib
Quoted 6 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Quoted 6 lines
Quoted 4 lines

Or changed to work only for node_type == e.IBV_NODE_CA?

Thanks,
Kamal

BTW, do the efa care about pkey?

Depends.. We only support the default pkey so it doesn't do much in terms of functionality, but we still need to support it as part of the QP state machine for modify QP.

#5 lore raw
Quoted 12 lines · Gal Pressman
Quoted 11 lines · Kamal Heib
Quoted 10 lines · Kamal Heib
Quoted 6 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Or changed to work only for node_type == e.IBV_NODE_CA?

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

Thanks,
Kamal

BTW, do the efa care about pkey?

Depends.. We only support the default pkey so it doesn't do much in terms of functionality, but we still need to support it as part of the QP state machine for modify QP.

Does the pkey appear on the wire, or is it just some cruft for API sake?

Jason

#6 lore raw
Quoted 20 lines · Jason Gunthorpe
Quoted 15 lines · Gal Pressman
Quoted 14 lines · Kamal Heib
Quoted 12 lines · Kamal Heib
Quoted 6 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Or changed to work only for node_type == e.IBV_NODE_CA?

Thanks,
Kamal

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

BTW, do the efa care about pkey?

Depends.. We only support the default pkey so it doesn't do much in terms of functionality, but we still need to support it as part of the QP state machine for modify QP.

Does the pkey appear on the wire, or is it just some cruft for API sake?

On the wire. Included in the BTH (Base Transfer Header).

Håkon

Quoted 2 lines

Jason

#7 lore raw
Quoted 19 lines · Håkon Bugge
Quoted 19 lines · Jason Gunthorpe
Quoted 17 lines · Gal Pressman
Quoted 15 lines · Kamal Heib
Quoted 5 lines · Kamal Heib
Quoted 4 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Quoted 7 lines
Quoted 2 lines

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Or changed to work only for node_type == e.IBV_NODE_CA?

Thanks,
Kamal

BTW, do the efa care about pkey?

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

Depends.. We only support the default pkey so it doesn't do much in terms of functionality, but we still need to support it as part of the QP state machine for modify QP.

Does the pkey appear on the wire, or is it just some cruft for API sake?

On the wire. Included in the BTH (Base Transfer Header).

He was probably asking specifically about EFA.
I can't share any details about the wire protocol, does it matter?

#8 lore raw
Quoted 24 lines · Gal Pressman
Quoted 23 lines · Håkon Bugge
Quoted 23 lines · Jason Gunthorpe
Quoted 20 lines · Gal Pressman
Quoted 9 lines · Kamal Heib
Quoted 8 lines · Kamal Heib
Quoted 6 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Quoted 6 lines
Quoted 4 lines

Or changed to work only for node_type == e.IBV_NODE_CA?

Thanks,
Kamal

BTW, do the efa care about pkey?

Depends.. We only support the default pkey so it doesn't do much in terms of functionality, but we still need to support it as part of the QP state machine for modify QP.

Quoted 6 lines
Quoted 3 lines
Quoted 1 line

Does the pkey appear on the wire, or is it just some cruft for API sake?

On the wire. Included in the BTH (Base Transfer Header).

He was probably asking specifically about EFA.
I can't share any details about the wire protocol, does it matter?

If it isn't actually used for anything then the driver shouldn't expose PKEY at all, if you do use it then leave it.

Jason

#9 lore raw
Quoted 27 lines · Jason Gunthorpe
Quoted 26 lines · Gal Pressman
Quoted 25 lines · Håkon Bugge
Quoted 24 lines · Jason Gunthorpe
Quoted 12 lines · Gal Pressman
Quoted 11 lines · Kamal Heib
Quoted 10 lines · Kamal Heib
Quoted 6 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Or changed to work only for node_type == e.IBV_NODE_CA?

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

Thanks,
Kamal

BTW, do the efa care about pkey?

Depends.. We only support the default pkey so it doesn't do much in terms of functionality, but we still need to support it as part of the QP state machine for modify QP.

Does the pkey appear on the wire, or is it just some cruft for API sake?

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

On the wire. Included in the BTH (Base Transfer Header).

He was probably asking specifically about EFA.
I can't share any details about the wire protocol, does it matter?

If it isn't actually used for anything then the driver shouldn't expose PKEY at all, if you do use it then leave it.

How would that work? How can you not expose pkeys and still have ibv_ud_pingpong work?
You mean remove query_pkey, but still support IBV_QP_PKEY_INDEX in modify_qp?

#10 lore raw
Quoted 33 lines · Gal Pressman
Quoted 32 lines · Jason Gunthorpe
Quoted 30 lines · Gal Pressman
Quoted 19 lines · Håkon Bugge
Quoted 19 lines · Jason Gunthorpe
Quoted 17 lines · Gal Pressman
Quoted 15 lines · Kamal Heib
Quoted 5 lines · Kamal Heib
Quoted 4 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Quoted 7 lines
Quoted 2 lines

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Or changed to work only for node_type == e.IBV_NODE_CA?

Thanks,
Kamal

BTW, do the efa care about pkey?

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

Depends.. We only support the default pkey so it doesn't do much in terms of functionality, but we still need to support it as part of the QP state machine for modify QP.

Does the pkey appear on the wire, or is it just some cruft for API sake?

On the wire. Included in the BTH (Base Transfer Header).

He was probably asking specifically about EFA.
I can't share any details about the wire protocol, does it matter?

Quoted 6 lines
Quoted 2 lines

If it isn't actually used for anything then the driver shouldn't expose PKEY at all, if you do use it then leave it.

How would that work? How can you not expose pkeys and still have ibv_ud_pingpong work?
You mean remove query_pkey, but still support IBV_QP_PKEY_INDEX in modify_qp?

iWarp doesn't use PKEY at all, so you'd have to follow their model. Some of the userspace that assumes IB would probably need fixing (eg ud pingpong does not support iwarp)

Jason

#11 lore raw
Quoted 37 lines · Jason Gunthorpe
Quoted 36 lines · Gal Pressman
Quoted 34 lines · Jason Gunthorpe
Quoted 24 lines · Gal Pressman
Quoted 23 lines · Håkon Bugge
Quoted 23 lines · Jason Gunthorpe
Quoted 20 lines · Gal Pressman
Quoted 9 lines · Kamal Heib
Quoted 8 lines · Kamal Heib
Quoted 6 lines · Gal Pressman
Quoted 2 lines · Kamal Heib

Now that the query_pkey() isn't mandatory by the RDMA core, this callback can be removed from the usnic provider.

Not directly related to this patch, but pyverbs has a test which verifies that max_pkeys > 0, maybe this check should be removed.

Quoted 6 lines
Quoted 4 lines

Or changed to work only for node_type == e.IBV_NODE_CA?

Thanks,
Kamal

BTW, do the efa care about pkey?

Depends.. We only support the default pkey so it doesn't do much in terms of functionality, but we still need to support it as part of the QP state machine for modify QP.

Quoted 6 lines
Quoted 3 lines
Quoted 1 line

Does the pkey appear on the wire, or is it just some cruft for API sake?

On the wire. Included in the BTH (Base Transfer Header).

He was probably asking specifically about EFA.
I can't share any details about the wire protocol, does it matter?

If it isn't actually used for anything then the driver shouldn't expose PKEY at all, if you do use it then leave it.

Quoted 7 lines
Quoted 3 lines

How would that work? How can you not expose pkeys and still have ibv_ud_pingpong work?
You mean remove query_pkey, but still support IBV_QP_PKEY_INDEX in modify_qp?

iWarp doesn't use PKEY at all, so you'd have to follow their model. Some of the userspace that assumes IB would probably need fixing (eg ud pingpong does not support iwarp)

Thanks, I'd rather keep supporting the default pkey only.